diff --git a/examples/evaluating_model_fit.ipynb b/examples/evaluating_model_fit.ipynb new file mode 100644 index 0000000..693342b --- /dev/null +++ b/examples/evaluating_model_fit.ipynb @@ -0,0 +1,185 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "73a88bcb-ffcb-4b96-90e4-b8422c627f3c", + "metadata": {}, + "source": [ + "# Evaluating the IMPALA Model Fit\n", + "In this document, we introduce some IMPALA functionality for evaluating the quality of the IMPALA calibration in terms of MCMC diagnostics/convergence, prediction in the training data, and statistical coverage of corresponding uncertainties. This is a non-exhaustive set of analyses that one can use to diagnose IMPALA outputs, but it should provide a solid foundational toolkit. We refer interested users to our other tutories (e.g., troubleshooting_convergence) for additional details on specific issues." + ] + }, + { + "cell_type": "markdown", + "id": "c205167c", + "metadata": {}, + "source": [ + "\n", + "## 1. Diagnosing the MCMC Sampler\n", + "IMPALA uses Markov Chain Monte Carlo sampling for parameter estimation, and users should do at minimum some basic evaluation of the performance of the sampler. Additional details are provided elsewhere; here, we just introduce some default functionality available to users with some basic visualization. " + ] + }, + { + "cell_type": "markdown", + "id": "c8725fae", + "metadata": {}, + "source": [ + "### Trace Plots\n", + "Trace plots show the posterior draws for various parameters (y-axis) as a function of the number of iterations of the MCMC sampler (x-axis). Visual evaluation of these plots provides a sense of how well the sampler has converged, how correlated parameter draws are across iterations and between different parameters, how many iterations to toss out at the start of the MCMC chain as burn-in, etc. \n", + "\n", + "Here is an example of code one might run to plot to theta0 draws for a hierarchical calibration stored in *out*:\n", + "\n", + " import impala.superCal.post_process as pp\n", + " pp.parameter_trace_plot(out.theta0[:,0,:], ylim=[0, 1])\n", + "\n", + "Here is an example trace plot. Notice the high autocorrelation evident in the last two parameters. In this calibration, the last two parameters were poorly identified with essentially uniform posteriors.\n", + "![something](./images/trace_example.png)" + ] + }, + { + "cell_type": "markdown", + "id": "4ad5c758", + "metadata": {}, + "source": [ + "### Performance of Parallel Tempering\n", + "IMPALA uses a sophisticated parameter sampling method called parallel tempering, which allows the sampler to more easily move between local modes of the posterior distribution. Parallel tempering works by running many parallel MCMC chains with different properties and at each iteration proposing swaps between chains. One useful diagnostic of the health of the parallel tempering is the proportion of iterations in which a proposed swap between parallel chains is accepted. Ideally, we would like this acceptance rate somewhere between 20-60%. \n", + "\n", + "For pooled and hierarchical calibrations, a diagnostic plot can be generated for a calibration stored in *out* and a IMPALA setup object stored in *setup*:\n", + "\n", + " pp.total_temperature_swaps(out,setup)\n", + "\n", + "Here is an example output plot. The total proportion of swaps accepted is shown in the text, and the proportion of swaps between pairs of chains (indexed by temperatures) is shown in a heatmap. This calibration shows some evidence of sub-optimal tempering performance. \n", + "\n", + "![something](./images/tempering_example.png)" + ] + }, + { + "cell_type": "markdown", + "id": "9679a3d3", + "metadata": {}, + "source": [] + }, + { + "cell_type": "markdown", + "id": "44be0aeb", + "metadata": {}, + "source": [ + "### Additional Convergence Diagnostics\n", + "Trace plots and proportion of parallel swaps accepted are the minimal set of convergence diagnostics one might want to run. However, many more diagnostics exist. The pyMC module in Python provides some useful functions for evaluating convergence (e.g., https://pymcmc.readthedocs.io/en/latest/modelchecking.html), and these methods can be applied to many objects stored in the outputted IMPALA calibration object to evaluate convergence." + ] + }, + { + "cell_type": "markdown", + "id": "b7aa21da", + "metadata": {}, + "source": [ + "## 2. Plotting the Posterior Draws for Theta\n", + "\n", + "### Pairs Plots\n", + "Pairs plots show parameter draws for pairs of parameters. A standard pairs plot can be generated using the *pairs* function, and specialized pairs plots for each calibration type can also be generated. \n", + "\n", + " pp.pairs(...)\n", + " pp.pairwise_theta_plot_pool(...)\n", + " pp.pairwise_theta_plot_hier(...)\n", + " pp.pairwise_theta_plot_cluster(...)\n", + "\n", + "Here is an example of a plot generated by *pairwise_theta_plot_hier*. Parameter theta_i indicates experiment-specific parameter draws, theta_0 indicates the mean of the parent distribution for theta_i's, and theta_star indicates the parent distribution itself (i.e., theta_0 draws + additional noise). \n", + "\n", + "![something](./images/pairs_hier_example.png)\n", + "\n", + "### Experiment-Specific Parameter Boxplots (Hierarchical and Clustering Only)\n", + "\n", + "Suppose we have a setup object with a single *addVecExperiments* call and that nexp is the number of unique theta_i value/experiments entered by the single call. Here is some code that can be used to generate boxplots of the theta_i values across experiments. This type of plot is useful for identifying how variable the theta_i values are across experiments. \n", + "\n", + " import numpy as np\n", + " import pandas as pd\n", + " import seaborn as sns\n", + " KEYS = np.array(pd.DataFrame(setup.bounds.keys())).flatten()\n", + " theta_i = []\n", + " for j in range(nexp):\n", + " mat = pd.DataFrame(np.array(pd.DataFrame(sc.tran_unif(out.theta[0][:,0,j,:],setup.bounds_mat, setup.bounds.keys()).values())).T, columns = KEYS)\n", + " mat['exp'] = j\n", + " if j == 0:\n", + " theta_i = [mat]\n", + " else:\n", + " theta_i.append(mat)\n", + " theta_i_long = pd.concat(theta_i)\n", + "\n", + " fig,ax=plt.subplots(1,setup.p,figsize=(20,4))\n", + " fig.tight_layout(pad=2.0)\n", + " for j in range(setup.p):\n", + " ax[j].set_xlabel('Experiment')\n", + " ax[j].set_ylabel('')\n", + " ax[j].set_title(KEYS[j]) \n", + " A = sns.boxplot(data = theta_i_long, x = theta_i_long['exp'], y = theta_i_long[KEYS[j]], ax=ax[j],showfliers=False)\n", + " A.set(xlabel='Experiment',ylabel='')\n", + " A.set(xticklabels=[]) \n", + "\n", + "Here is an example output: \n", + "![something](./images/thetai_example.png)\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "id": "33249e09", + "metadata": {}, + "source": [ + "## 3. Evaluating Predictions in Training Data\n", + "Another important diagnostic is how well the calibrated model predicts the training data. Poor predictions can indicate model discrepancy, errors in implementation, etc. IMPALA provides several utilities for calculating and visualizing predictions. \n", + "\n", + "### Getting and Plotting Predictions\n", + "The function *get_outcome_predictions_impala* can be used to generate predictions for all training data points and a user-specified set of theta values. The *post_process.py* script also has various other functions that help with plotting these predictions. \n", + "\n", + "For example, prediction plots specific to ModelMaterialStrength calibrations can be generated as follows:\n", + "\n", + " pp.ptw_prediction_plots_pool(...)\n", + " pp.ptw_prediction_plots_hier(...)\n", + " pp.ptw_prediction_plots_cluster(...)\n", + "\n", + "Generate prediction plots can be generated using the *func_prediction_plot* function. \n", + "\n", + "### Cross-Experiment Prediction Errors (Hierarchical and Clustering Only)\n", + "\n", + "One extremely useful diagnostic for identifying outliers and determining if there are subsets of experiments that behave \"differently\" than others is to calculate the predictions for experiment i using the experiment-specific theta values estimated for other experiments. If an experiment has low prediction error using its own theta_i estimate but high prediction error using other experiments' theta estimates, there is some evidence that the experiment is somehow \"different\" than the others. \n", + "\n", + "Here, we provide some code for generating these prediction errors in a particular setting. This code can be adapted by users to accomodate different model structures. For this demonstration, we suppose we have a setup object with a single *addVecExperiments* call and that nexp is the number of unique theta_i value/experiments entered by the single call. \n", + "\n", + " THETAi_Y = [get_outcom_predictions_impala(setup, theta_input = np.array(pd.DataFrame(sc.tran_unif(out.theta[0][:,0,j,:],setup.bounds_mat, setup.bounds.keys()).values())).T)['outcome_draws'] for j in range(nexp)]\n", + " CROSS_ERRORS = np.empty([nexp,nexp])\n", + " for i in range(nexp): \n", + " for j in range(nexp):\n", + " CROSS_ERRORS[i,j] = np.mean(np.abs(setup.ys[0][np.where(setup.s2_ind[0] == j)[0]] -np.mean(THETAi_Y[i][0][:,np.where(setup.s2_ind[0] == j)[0]],axis=0))/setup.ys[0][np.where(setup.s2_ind[0] == j)[0]])\n", + " ax = sns.heatmap(CROSS_ERRORS.T*100, linewidths =0)\n", + " ax.set_xlabel('Theta Index')\n", + " ax.set_ylabel('Prediction Index')\n", + " ax.set_title('Prediction Errors (%)')\n", + "\n", + "Here, we provide an example output. Rows correspond to experiments being predicted and columns correspond to experiment-specific theta_i values. The color in each cell corresponds to the mean absolute percent error (MAPE) for predicting the training data. This figure indicates some differing behavior between the first 7 and last 9 experiments. It also highlights experiment 6 as particularly poorly-predicted using the theta_i values obtained for experiments 7-15. \n", + "\n", + "![something](./images/cross_errors_example.png)\n" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.8.18" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/examples/ex_bayarri_discrepancy.ipynb b/examples/ex_bayarri_discrepancy.ipynb index dba96b6..c4b36de 100644 --- a/examples/ex_bayarri_discrepancy.ipynb +++ b/examples/ex_bayarri_discrepancy.ipynb @@ -8,7 +8,7 @@ "### Example Calibration with Discrepancy and Emulation - Bayarri function\n", "In this example, we calibrate a modification of the function explored in Bayarri et al. (2007) titled \"A framework for validation of computer models\" in Technometrics incorporating discrepancy.\n", "\n", - "Calibrations involving discrepancy tend to be challenging, since there is often at best weak identifiability between input parameters and parameters in the discrepancy model. By \"weak identifiability\", we mean that there are This is a difficult calibration problem there are many combinations of inputs that produce similar fits to the training data. In this example, we will also demonstrate simulation emulation using the pyBASS package." + "Calibrations involving discrepancy tend to be challenging, since there is often at best weak identifiability between input parameters and parameters in the discrepancy model. By \"weak identifiability\", we mean that there are This is a difficult calibration problem there are many combinations of inputs that produce similar fits to the training data. In this example, we will also demonstrate simulation emulation using the **pyBASS** package available at https://github.com/lanl/pyBASS." ] }, { @@ -119,7 +119,7 @@ "id": "b40c6656", "metadata": {}, "source": [ - "Now, we suppose that the functional form of g(x) is not available and must be emulated. We will use the bassPCA method from the pyBASS package to do the emulation. " + "Now, we suppose that the functional form of g(x) is not available and must be emulated. We will use the **bassPCA** method from the **pyBASS** package to do the emulation. " ] }, { @@ -185,7 +185,7 @@ "id": "94dad9c6", "metadata": {}, "source": [ - "Now, we define the IMPALA model structure. Since we used bassPCA to model this functional data, we use the following:" + "Now, we define the IMPALA model structure. Since we used **bassPCA** to model this functional data, we use the following:" ] }, { @@ -265,7 +265,7 @@ "id": "e1811d6a-6462-41ea-acdc-d43b32d502a4", "metadata": {}, "source": [ - "We can look at trace plots of the calibration parameters (theta) for the models with and withour discrepancy. We want these to look like they converge, and choose which MCMC iterations to use accordingly." + "We can look at trace plots of the calibration parameters (theta) for the models with and without discrepancy. We want these to look like they converge, and choose which MCMC iterations to use accordingly." ] }, { @@ -603,7 +603,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.8.16" + "version": "3.8.18" } }, "nbformat": 4, diff --git a/examples/ex_friedman.ipynb b/examples/ex_friedman.ipynb index cbc5747..e86cb28 100644 --- a/examples/ex_friedman.ipynb +++ b/examples/ex_friedman.ipynb @@ -104,7 +104,7 @@ "id": "ccdce893-9ff8-4bbf-a2e1-bf8b1a7cf499", "metadata": {}, "source": [ - "Now we are ready to set up and run our calibration. Since the functional form of f(x) is assumed known and fast to compute, we specify the model to calibrate using the ModelF class." + "Now we are ready to set up and run our calibration. Since the functional form of f(x) is assumed known and fast to compute, we specify the model to calibrate using the **ModelF** class." ] }, { @@ -363,7 +363,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.8.16" + "version": "3.8.18" } }, "nbformat": 4, diff --git a/examples/ex_shpb_clustering.ipynb b/examples/ex_shpb_clustering.ipynb index 8971962..eb5698a 100644 --- a/examples/ex_shpb_clustering.ipynb +++ b/examples/ex_shpb_clustering.ipynb @@ -266,9 +266,9 @@ "id": "ccdce893-9ff8-4bbf-a2e1-bf8b1a7cf499", "metadata": {}, "source": [ - "Now we are ready to set up and run our calibration. If several distinct types of experiments (e.g., SHPB and flyer plate experiments) are being used, additional experiments and models can be added to \"setup\" by calling setup.addVecExperiments multiple times.\n", + "Now we are ready to set up and run our calibration. If several distinct types of experiments (e.g., SHPB and flyer plate experiments) are being used, additional experiments and models can be added to \"setup\" by calling **setup.addVecExperiments** multiple times.\n", "\n", - "This code runs a clustered calibration analysis with the following model structure: yobs ~ Normal( PTW(theta_e)) , s2_e ), where theta_e takes one of nclustmax values based on the experiment's cluster. Parameter s2_e is an experiment-specific noise parameter. Letting omega_g index nclustmax theta values, assume omega_g ~ DirichletProcess(G0, eta), where G0 is a Normal(theta_0, Sigma_0) distribution for some shared mean theta_0 and covariance Sigma_0 and where eta is a concentration parameter controlling the propensity for experiments to join existing clusters. Larger values of alpha are associated with less concentration of experiments into shared clusters. Here, PTW(.) indicates a PTW model evaluation. \n", + "This code runs a clustered calibration analysis with the following model structure: yobs ~ Normal( PTW(theta_e)) , s2_e ), where theta_e takes one of **nclustmax** values based on the experiment's cluster. Parameter s2_e is an experiment-specific noise parameter. Letting omega_g index **nclustmax** theta values, assume omega_g ~ DirichletProcess(G0, eta), where G0 is a Normal(theta_0, Sigma_0) distribution for some shared mean theta_0 and covariance Sigma_0 and where eta is a concentration parameter controlling the propensity for experiments to join existing clusters. Larger values of alpha are associated with less concentration of experiments into shared clusters. Here, PTW(.) indicates a PTW model evaluation. \n", "\n", "For additional details on this Dirichlet Process model specification, we refer interested readers to \"Markov Chain Sampling Methods for Dirichlet Process Mixture Models\" by Radford M Neal in the Journal of Computational and Graphical Statistics; 9(2); 2000." ] @@ -784,7 +784,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.8.16" + "version": "3.8.18" } }, "nbformat": 4, diff --git a/examples/ex_shpb_hierarchical.ipynb b/examples/ex_shpb_hierarchical.ipynb index af9896d..c4e19e0 100644 --- a/examples/ex_shpb_hierarchical.ipynb +++ b/examples/ex_shpb_hierarchical.ipynb @@ -266,7 +266,7 @@ "id": "ccdce893-9ff8-4bbf-a2e1-bf8b1a7cf499", "metadata": {}, "source": [ - "Now we are ready to set up and run our calibration. If several distinct types of experiments (e.g., SHPB and flyer plate experiments) are being used, additional experiments and models can be added to \"setup\" by calling setup.addVecExperiments multiple times.\n", + "Now we are ready to set up and run our calibration. If several distinct types of experiments (e.g., SHPB and flyer plate experiments) are being used, additional experiments and models can be added to \"setup\" by calling **setup.addVecExperiments** multiple times.\n", "\n", "This code runs a hierarchical calibration analysis with the following model structure: yobs ~ Normal( PTW(theta_e)) , s2_e ), where theta_e and s2_e are experiment-specific parameters and noise values and where we assume theta_e ~ Normal(theta_0, Sigma_0) for some shared mean theta_0 and covariance Sigma_0. Here, PTW(.) indicates a PTW model evaluation." ] @@ -693,7 +693,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.8.16" + "version": "3.8.18" } }, "nbformat": 4, diff --git a/examples/ex_shpb_pooled.ipynb b/examples/ex_shpb_pooled.ipynb index 8686fc9..5f86f22 100644 --- a/examples/ex_shpb_pooled.ipynb +++ b/examples/ex_shpb_pooled.ipynb @@ -266,7 +266,7 @@ "id": "ccdce893-9ff8-4bbf-a2e1-bf8b1a7cf499", "metadata": {}, "source": [ - "Now we are ready to set up and run our calibration. If several distinct types of experiments (e.g., SHPB and flyer plate experiments) are being used, additional experiments and models can be added to \"setup\" by calling setup.addVecExperiments multiple times.\n", + "Now we are ready to set up and run our calibration. If several distinct types of experiments (e.g., SHPB and flyer plate experiments) are being used, additional experiments and models can be added to \"setup\" by calling **setup.addVecExperiments** multiple times.\n", "\n", "This code runs a pooled calibration analysis with the following model structure: yobs ~ Normal( PTW(theta) , s2 ), where theta is shared across all experiemnts and PTW(.) indicates a PTW model evaluation." ] @@ -531,7 +531,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.8.16" + "version": "3.8.18" } }, "nbformat": 4, diff --git a/examples/images/cross_errors_example.png b/examples/images/cross_errors_example.png new file mode 100644 index 0000000..1de8304 Binary files /dev/null and b/examples/images/cross_errors_example.png differ diff --git a/examples/images/pairs_hier_example.png b/examples/images/pairs_hier_example.png new file mode 100644 index 0000000..3e19040 Binary files /dev/null and b/examples/images/pairs_hier_example.png differ diff --git a/examples/images/tempering_example.png b/examples/images/tempering_example.png new file mode 100644 index 0000000..8dd01ff Binary files /dev/null and b/examples/images/tempering_example.png differ diff --git a/examples/images/thetai_example.png b/examples/images/thetai_example.png new file mode 100644 index 0000000..a68fdbb Binary files /dev/null and b/examples/images/thetai_example.png differ diff --git a/examples/images/trace_example.png b/examples/images/trace_example.png new file mode 100644 index 0000000..b723b40 Binary files /dev/null and b/examples/images/trace_example.png differ diff --git a/examples/impala_options.ipynb b/examples/impala_options.ipynb index 94a4945..ceb4c29 100644 --- a/examples/impala_options.ipynb +++ b/examples/impala_options.ipynb @@ -22,48 +22,47 @@ "\n", "### Initialize Impala Model\n", "\n", - "1. ModelMaterialStrength: class for pre-defined IMPALA material strength models\n", + "1. **ModelMaterialStrength**: class for pre-defined IMPALA material strength models\n", "\n", - "2. ModelBassPca_func or ModelBassPca_mult: classes for functional or multivariate emulators fit using the pyBASS library\n", + "2. **ModelBassPca_func** or **ModelBassPca_mult**: classes for functional or multivariate emulators fit using the pyBASS library\n", " \n", - "3. ModelF: class for user-defined functions f(x), which can also be used to implement IMPALA with non-BASS emulators\n", + "3. **ModelF**: class for user-defined functions f(x), which can also be used to implement IMPALA with non-BASS emulators\n", "\n", "### Prepare the Fit\n", "\n", - "1. CalibSetup: initializes an IMPALA calibration object. This is also where you specify parameter bounds and any constraints\n", + "1. **CalibSetup**: initializes an IMPALA calibration object. This is also where you specify parameter bounds and any constraints\n", "\n", - "2. addVecExperiments: define the observed data, corresponding computer model, discrepancy basis (if any), and several noise model prior hyperparameters. Multiple addVecExperiments calls can be used to add different experiments, possibly with different corresponding computer models. Some inputs include:\n", - " * yobs: a vector (numpy array) of observed data\n", - " * model: an IMPALA model object as defined above. See code documentation for details. \n", - " * sd_est: a list or numpy array of initial values for observation noise standard deviation\n", - " * s2_df: a list or numpy array of initial values for s2 Inverse Gamma prior degrees of freedom\n", - " * s2_ind: a list or numpy array of indices for s2 value associated with each element of yobs\n", - " * meas_error_cor: (optional) correlation matrix for observation measurement errors, default = independent \n", - " * D: (optional) numpy array containing basis functions for discrepancy, possibly including intercept.\n", - " * discrep_tau: (optional) fixed prior variance for discrepancy basis coefficients (discrepancy = D @ discrep_vars, \n", - " * discrep_vars ~ N(0,discrep_tau))\n", + "2. **addVecExperiments**: define the observed data, corresponding computer model, discrepancy basis (if any), and several noise model prior hyperparameters. Multiple addVecExperiments calls can be used to add different experiments, possibly with different corresponding computer models. Some inputs include:\n", + " * *yobs*: a vector (numpy array) of observed data\n", + " * *model*: an IMPALA model object as defined above. See code documentation for details. \n", + " * *sd_est*: a list or numpy array of initial values for observation noise standard deviation\n", + " * *s2_df*: a list or numpy array of initial values for s2 Inverse Gamma prior degrees of freedom\n", + " * *s2_ind*: a list or numpy array of indices for s2 value associated with each element of yobs\n", + " * *meas_error_cor*: (optional) correlation matrix for observation measurement errors, default = independent \n", + " * *D*: (optional) numpy array containing basis functions for discrepancy, possibly including intercept.\n", + " * *discrep_tau*: (optional) fixed prior variance for discrepancy basis coefficients (discrepancy = D @ discrep_vars, discrep_vars ~ N(0,discrep_tau))\n", "\n", - "3. setTemperatureLadder: define how the parallel tempering should be implemented, requiring users to specify an array of exponents that will be applied to the data likelihood. An example specifcation is np.array(1.05 ** np.linspace(0,49,50)), which assigns a grid of 50 temperatures. Generally, more temperatures or a finer grid of temperatures are associated with longer runtime but may also be associated with better movement around the posterior surface for complicated posteriors. \n", + "3. **setTemperatureLadder**: define how the parallel tempering should be implemented, requiring users to specify an array of exponents that will be applied to the data likelihood. An example specifcation is np.array(1.05 ** np.linspace(0,49,50)), which assigns a grid of 50 temperatures. Generally, more temperatures or a finer grid of temperatures are associated with longer runtime but may also be associated with better movement around the posterior surface for complicated posteriors. \n", "\n", - "4. setMCMC: define how many MCMC iterations to use for the sampler. Most users can leave these settings at default values, with the expection of nmcmc (the number of iterations), which must be specified. \n", + "4. **setMCMC**: define how many MCMC iterations to use for the sampler. Most users can leave these settings at default values, with the expection of nmcmc (the number of iterations), which must be specified. \n", "\n", - "5. setHierPriors: (optional) define hyperparameteters associated with the hierarchical and clustering calibrations. These generally control the amount of shrinkage toward a common theta across experiments. Please refer to the code documnetation for details. \n", + "5. **setHierPriors**: (optional) define hyperparameteters associated with the hierarchical and clustering calibrations. These generally control the amount of shrinkage toward a common theta across experiments. Please refer to the code documnetation for details. \n", "\n", - "6. setClusterPriors: (optional) define hyperparameteters associated with the clustering calibration, including the maximum number of clusters (nclustmax) and the rate and shape associated with the Gamma prior on the Dirichlet process concentration parameter, eta. \n", + "6. **setClusterPrior**: (optional) define hyperparameteters associated with the clustering calibration, including the maximum number of clusters (nclustmax) and the rate and shape associated with the Gamma prior on the Dirichlet process concentration parameter, eta. \n", "\n", "### Run MCMC\n", "\n", - "1. calibPool: pooled calibration\n", + "1. **calibPool**: pooled calibration\n", "\n", - "2. calibHier: hierarchical calibration\n", + "2. **calibHier**: hierarchical calibration\n", "\n", - "3. calibClust: clustered calibration\n", + "3. **calibClust**: clustered calibration\n", "\n", "### Evaluate Convergence\n", "\n", - "1. Look at trace plots, e.g., using parameter_trace_plot function\n", + "1. Look at trace plots, e.g., using **parameter_trace_plot** function\n", "\n", - "2. Look at pairs plots, e.g., using pairs function\n", + "2. Look at pairs plots, e.g., using **pair**s function\n", "\n", "3. There are many more convergence diagnostics you can explore. In the future, additional convergence evaluation functions will be added to the IMPALA repo. \n", "\n", diff --git a/examples/material_strength.ipynb b/examples/material_strength.ipynb new file mode 100644 index 0000000..0d37ffc --- /dev/null +++ b/examples/material_strength.ipynb @@ -0,0 +1,140 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "73a88bcb-ffcb-4b96-90e4-b8422c627f3c", + "metadata": {}, + "source": [ + "# Calibrating Material Strength Models with IMPALA\n", + "This document summarizes the current IMPALA functionality for calibrating material strength models using the model class **ModelMaterialStrength** for Hopkinson Bar and Quasistatic experiments. This class assumes the observed data for a given experiment consist of an initial temperature (Kelvin), a strain rate (per microseconds), and vectors with strains (unitless) and stress values (megabars).\n", + "\n", + "**Note:** We provide suggested uits for each of the inputs, but the code is written such that users can enter data and corresponding model constants in whatever units they want, provided that the units are all internally consistent. " + ] + }, + { + "cell_type": "markdown", + "id": "317605ac", + "metadata": {}, + "source": [ + "\n", + "### About ModelMaterialStrength\n", + "\n", + "Models of class **ModelMaterialStrength** are defined using the following physics inputs:\n", + "\n", + "1. **temps** : a list of temperatures indexed by experiment (Kelvin)\n", + "2. **edots** : list of strain rates indexed by experiment (per microsecond)\n", + "3. **consts** : a dictionary of constants for the material stress model. See next section for more details. \n", + "4. **strain_histories** : a list of strain histories for Hopkinson Bar/Quasistatic experiments (unitless between 0 and 1, relative to starting length)\n", + "5. **flow_stress_model** : model for flow/yield stress. Options include constant, Johnson-Cook, Preston-Tonks-Wallance, and Stein\n", + "6. **melt_model** : model for melt temperature. \n", + "7. **shear_model** : model for shear modulus\n", + "8. **specific_heat_model** : model for specific heat\n", + "9. **density_model** : model for density\n", + "\n", + "When the calibration model is set up, stress observations are in terms of megabars. " + ] + }, + { + "cell_type": "markdown", + "id": "dd680084", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "source": [ + "### Listing Options for Submodels\n", + "\n", + "Many different flow stress, melt, shear, specific heat, and density models have been implemented, and users can specify which of models that make sense for their material of interest. Each specified model is associated with additional constants, which are specified in the consts dictionary in the **ModelMaterialStrength** call. Alternatively, users can specify their own functions for these different models. These functions should follow the input/output structure of the options already implemented in IMPALA, and corresponding constants should be added on the constants dictionary in **ModelMaterialStrength**. \n", + "\n", + "Below, we list the options currently implemented for each model. The model structure and corresponding constants can be accessed using **showdef_ModelMaterialStrength(str_model_name)**." + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "35fb3b5b", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "['Constant_Yield_Stress', 'JC_Yield_Stress', 'PTW_Yield_Stress', 'Stein_Flow_Stress']\n", + "['BGP_Melt_Temperature', 'Constant_Melt_Temperature', 'Linear_Melt_Temperature', 'Quadratic_Melt_Temperature']\n", + "['BGP_PW_Shear_Modulus', 'Constant_Shear_Modulus', 'Linear_Cold_PW_Shear_Modulus', 'Quadratic_Cold_PW_Shear_Modulus', 'Simple_Shear_Modulus', 'Stein_Shear_Modulus']\n", + "['Constant_Specific_Heat', 'Linear_Specific_Heat', 'Piecewise_Linear_Specific_Heat', 'Piecewise_Quadratic_Specific_Heat', 'Quadratic_Specific_Heat']\n", + "['Constant_Density', 'Cubic_Density', 'Linear_Density', 'Quadratic_Density']\n" + ] + } + ], + "source": [ + "import impala.superCal.models_withLik as ml\n", + "print(ml.getoptions_ModelMaterialStrength()['flow_stress_model'])\n", + "print(ml.getoptions_ModelMaterialStrength()['melt_model'])\n", + "print(ml.getoptions_ModelMaterialStrength()['shear_model'])\n", + "print(ml.getoptions_ModelMaterialStrength()['specific_heat_model'])\n", + "print(ml.getoptions_ModelMaterialStrength()['density_model'])" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "956f1305", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "class Piecewise_Quadratic_Specific_Heat(BaseModel):\n", + " \"\"\"\n", + " Piecewise Quadratic Specific Heat Model\n", + " Cv (T) = c0_0 + c1_0 * T + c2_0 * T**2 for T<=T_t\n", + " Cv (T) = c0_1 + c1_1 * T + c2_1 * T**2 for T>T_t \n", + " \"\"\"\n", + " consts = ['T_t','c0_0', 'c1_0', 'c2_0', 'c0_1', 'c1_1', 'c2_1']\n", + " def value(self, *args):\n", + " tnow=self.parent.state.T\n", + " pow_0_coeff = np.repeat(self.parent.parameters.c0_0,len(tnow))\n", + " pow_1_coeff = np.repeat(self.parent.parameters.c1_0,len(tnow))\n", + " pow_2_coeff = np.repeat(self.parent.parameters.c2_0,len(tnow))\n", + " \n", + " pow_0_coeff[np.where(tnow > self.parent.parameters.T_t)] = self.parent.parameters.c0_1\n", + " pow_1_coeff[np.where(tnow > self.parent.parameters.T_t)] = self.parent.parameters.c1_1\n", + " pow_2_coeff[np.where(tnow > self.parent.parameters.T_t)] = self.parent.parameters.c2_1\n", + " \n", + " cnow = pow_0_coeff + pow_1_coeff * tnow + pow_2_coeff * tnow * tnow\n", + " return cnow\n", + "\n", + "None\n" + ] + } + ], + "source": [ + "print(ml.showdef_ModelMaterialStrength('Piecewise_Quadratic_Specific_Heat'))" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.8.18" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/examples/maximum_a_posteriori_estimation.ipynb b/examples/maximum_a_posteriori_estimation.ipynb new file mode 100644 index 0000000..17eb13a --- /dev/null +++ b/examples/maximum_a_posteriori_estimation.ipynb @@ -0,0 +1,517 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "73a88bcb-ffcb-4b96-90e4-b8422c627f3c", + "metadata": {}, + "source": [ + "# Maximum a Posteriori Estimation for Pooled Calibration with IMPALA\n", + "The main IMPALA codebase estimates calibration parameters theta using Bayesian Markov Chain Monte Carlo. This algorithm provides a posterior distribution for theta. This is more than we need, however, if we only care about the single \"best\" configuration of theta fitting our observed data. In that case, an often faster and more direct estimate of the \"best\" configuration of theta can be obtained using Maximum a Posteriori (MAP) estimation, a popular Bayesian approach to estimating the mode (i.e., highest point) of the posterior distribution.\n", + "\n", + "In as-yet unpublished work, we developed a MAP estimation algorithm for IMPALA that optimizes the posterior likelihood of the observed data as a function of theta, after integrating out nuisances parameters such as the experimental noise. We provide the corresponding code as part of IMPALA, which should be viewed as experimental for now. The integrated posterior is approximated using Monte Carlo Integration, an importance sampling-based methodology. The number of samples used for the Monte Carlo integration directly relates to the degree of approximation, where a larger number of samples is generally associated with a better approximation. In practice, we have found that 100 Monte Carlo samples provides a reasonable approximation in many settings. This approximated posterior is then maximized with respect to theta (and any discrepancy basis coefficients if applicable) using one of three different global maximization methods. Global maximization is essential for this problem, since calibration posterior distributions often have multiple local modes. We note that the posterior may not even have a unique local mode. Due to this, there are no guarantees of uniqueness for the resulting MAP estimate. \n", + "\n", + "Below, we describe how to implement the MAP estimation methods with IMPALA. Then, we illustrate the methods using the Friedman function. " + ] + }, + { + "cell_type": "markdown", + "id": "b7aa21da", + "metadata": {}, + "source": [ + "## IMPALA Code for MAP Estimation\n", + "The current implementation of the IMPALA MAP estimation uses the function **get_map_impalapool**. Arguments are as follows:\n", + "\n", + "1. **setup**: the IMPALA modeling object generated as usual (see other tutorials for details)\n", + "2. **n_samples**: number of Monte Carlo Integration samples used to approximate the integrated posterior. Generally, 100-1000 seem to do well. \n", + "3. **theta_init**: 1 x p array with the initial values for theta. This can be obtained using the IMPALA theta posterior as below or defined manually by the user.\n", + "4. **disc_init**: (optional) if the model includes discrepancy, this is a 1 x d array containing initial values for the discrepancy basis coefficients, where d is the number of basis elements.\n", + "5. **optmethod**: defines the global optimization algorithm used. Options include 'bh for basinhopping (default), 'pso' for particle swarm optimization, and 'grid' for optimization across Latin Hypercube grid values.\n", + "6. **niter**: number of iterations for basinhopping, swarmsize for particle swarm, and number of parallel runs for Latin Hypercube grid optimization\n", + "7. **T**: hyperparameter for basinhopping algorithm related to the size of the jumps between iterations. See https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.basinhopping.html for details. \n", + "8. **n_scores**: hyperparameter for Latin Hypercube sampling corresponding to number of cores used for parallel runs\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "id": "d0661f2a", + "metadata": {}, + "source": [ + "## An Illustrative Example: MAP Estimation for the Friedman Function\n", + "In this section, we illustrate the MAP estimation method for the Friedman function. Additional details about the Friedman function are provided in **ex_friedman**. Here, we run the example without additional exposition. \n", + "\n", + "### Generate Data and Initialize IMPALA Model Structure" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "5d2ab233", + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "from impala import superCal as sc\n", + "import impala.superCal.post_process as pp\n", + "import matplotlib.pyplot as plt\n", + "import pandas as pd\n", + "\n", + "### Generating Data \n", + "def f(x):\n", + " out = 10. * np.sin(np.pi * tt * x[0]) + 20. * (x[1] - .5) ** 2 + 10 * x[2] + 5. * x[3]\n", + " return out\n", + "p = 9\n", + "nt = 50\n", + "tt = np.linspace(0, 1, nt)\n", + "xx_true = np.random.rand(1, p)\n", + "yobs = np.apply_along_axis(f, 1, xx_true).reshape(nt) + np.random.normal(size=50)*.1\n", + "\n", + "### Setting up the IMPALA Model Structure\n", + "input_names = [str(v) for v in list(range(p))] \n", + "bounds = dict(zip(input_names, np.concatenate((np.zeros((p, 1)), np.ones((p, 1))), 1))) \n", + "setup = sc.CalibSetup(bounds, constraint_func='bounds') \n", + "model = sc.ModelF(f, input_names) \n", + "setup.addVecExperiments(yobs=yobs, \n", + " model=model, \n", + " sd_est=[1.], \n", + " s2_df=[0], \n", + " s2_ind=[0]*nt) \n", + "setup.setTemperatureLadder(1.05**np.arange(40)) \n", + "setup.setMCMC(nmcmc=15000,decor=100) \n", + "\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "id": "60104dc9", + "metadata": {}, + "source": [ + "### (Optional) Get good initial value for optimal theta\n", + "The following steps are optional. It is used to obtain a really good initialization for the MAP estimation algorithm. However, this step could be replaced with a user-defined initialization in the next step." + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "b45f9456", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[2024-05-03 09:34:32 | 14999/14999 (100%) | WALL: 0:00:38 | ETA: 0:00:00 | 388.44it/s]\n", + "Calibration MCMC Complete. Time: 38.625613 seconds.\n" + ] + } + ], + "source": [ + "### Run Pooled Calibration \n", + "out = sc.calibPool(setup) # pooled calibration (takes less than a minute)\n" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "16ae42d5", + "metadata": {}, + "outputs": [], + "source": [ + "### Identify Highest-Likelihood MCMC Sample of Theta\n", + "mcmc_use = np.arange(5000, 15000, 2) # burn and thin index\n", + "mat = np.vstack(out.theta[mcmc_use,0,:])\n", + "mat_trans = np.array(pd.DataFrame(sc.tran_unif( mat, setup.bounds_mat, setup.bounds.keys())))\n", + "POST_pool = sc.eval_partialintlogposterior_impalapool(setup, n_samples = 1000, theta = mat_trans)" + ] + }, + { + "cell_type": "markdown", + "id": "5af1467c", + "metadata": {}, + "source": [ + "### Obtain MAP Estimate\n", + "This function uses the default basinhopping algorithm with 100 iterations." + ] + }, + { + "cell_type": "code", + "execution_count": 42, + "id": "f32e072f", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "basinhopping step 0: f -81.8332\n", + "basinhopping step 1: f -81.8406 trial_f -81.8406 accepted 1 lowest_f -81.8406\n", + "found new global minimum on step 1 with function value -81.8406\n", + "basinhopping step 2: f -81.8406 trial_f -81.8406 accepted 1 lowest_f -81.8406\n", + "found new global minimum on step 2 with function value -81.8406\n", + "basinhopping step 3: f -81.8424 trial_f -81.8424 accepted 1 lowest_f -81.8424\n", + "found new global minimum on step 3 with function value -81.8424\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 4: f -81.8424 trial_f 1e+06 accepted 0 lowest_f -81.8424\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 5: f -81.8424 trial_f 1e+06 accepted 0 lowest_f -81.8424\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 6: f -81.8424 trial_f 1e+06 accepted 0 lowest_f -81.8424\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 7: f -81.8424 trial_f 1e+06 accepted 0 lowest_f -81.8424\n", + "basinhopping step 8: f -81.8428 trial_f -81.8428 accepted 1 lowest_f -81.8428\n", + "found new global minimum on step 8 with function value -81.8428\n", + "basinhopping step 9: f -81.8411 trial_f -81.8411 accepted 1 lowest_f -81.8428\n", + "basinhopping step 10: f -81.8429 trial_f -81.8429 accepted 1 lowest_f -81.8429\n", + "found new global minimum on step 10 with function value -81.8429\n", + "basinhopping step 11: f -81.835 trial_f -81.835 accepted 1 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 12: f -81.835 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 13: f -81.835 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 14: f -81.835 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 15: f -81.835 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "basinhopping step 16: f -81.835 trial_f 69.9125 accepted 0 lowest_f -81.8429\n", + "basinhopping step 17: f -81.8421 trial_f -81.8421 accepted 1 lowest_f -81.8429\n", + "basinhopping step 18: f -81.8365 trial_f -81.8365 accepted 1 lowest_f -81.8429\n", + "basinhopping step 19: f -81.8327 trial_f -81.8327 accepted 1 lowest_f -81.8429\n", + "basinhopping step 20: f -81.8396 trial_f -81.8396 accepted 1 lowest_f -81.8429\n", + "basinhopping step 21: f -81.8382 trial_f -81.8382 accepted 1 lowest_f -81.8429\n", + "basinhopping step 22: f -81.8365 trial_f -81.8365 accepted 1 lowest_f -81.8429\n", + "basinhopping step 23: f -81.8324 trial_f -81.8324 accepted 1 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 24: f -81.8324 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 25: f -81.8324 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 26: f -81.8324 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "basinhopping step 27: f -81.8324 trial_f 69.9125 accepted 0 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 28: f -81.8324 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 29: f -81.8324 trial_f 1e+06 accepted 0 lowest_f -81.8429\n", + "basinhopping step 30: f -81.8441 trial_f -81.8441 accepted 1 lowest_f -81.8441\n", + "found new global minimum on step 30 with function value -81.8441\n", + "basinhopping step 31: f -81.8413 trial_f -81.8413 accepted 1 lowest_f -81.8441\n", + "basinhopping step 32: f -81.8432 trial_f -81.8432 accepted 1 lowest_f -81.8441\n", + "basinhopping step 33: f -81.8378 trial_f -81.8378 accepted 1 lowest_f -81.8441\n", + "basinhopping step 34: f -81.8422 trial_f -81.8422 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 35: f -81.8422 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 36: f -81.8422 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 37: f -81.8422 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 38: f -81.8391 trial_f -81.8391 accepted 1 lowest_f -81.8441\n", + "basinhopping step 39: f -81.8391 trial_f 69.9125 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 40: f -81.8391 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 41: f -81.8391 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 42: f -81.8391 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 43: f -81.8419 trial_f -81.8419 accepted 1 lowest_f -81.8441\n", + "basinhopping step 44: f -81.835 trial_f -81.835 accepted 1 lowest_f -81.8441\n", + "basinhopping step 45: f -81.8431 trial_f -81.8431 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 46: f -81.8431 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 47: f -81.8431 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 48: f -81.8431 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 49: f -81.8431 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "adaptive stepsize: acceptance rate 0.460000 target 0.500000 new stepsize 0.09 old stepsize 0.1\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 50: f -81.8431 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 51: f -81.7618 trial_f -81.7618 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 52: f -81.7618 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 53: f -81.7618 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 54: f -81.7618 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 55: f -81.7618 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 56: f -81.7618 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 57: f -81.7618 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 58: f -81.8441 trial_f -81.8441 accepted 1 lowest_f -81.8441\n", + "basinhopping step 59: f -81.8211 trial_f -81.8211 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 60: f -81.8211 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 61: f -81.8342 trial_f -81.8342 accepted 1 lowest_f -81.8441\n", + "basinhopping step 62: f -81.8425 trial_f -81.8425 accepted 1 lowest_f -81.8441\n", + "basinhopping step 63: f -81.8383 trial_f -81.8383 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 64: f -81.8383 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 65: f -81.8408 trial_f -81.8408 accepted 1 lowest_f -81.8441\n", + "basinhopping step 66: f -81.8404 trial_f -81.8404 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 67: f -81.8404 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 68: f -81.8433 trial_f -81.8433 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 69: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 70: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 71: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 72: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 73: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 74: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 75: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 76: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 77: f -81.8433 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 78: f -81.835 trial_f -81.835 accepted 1 lowest_f -81.8441\n", + "basinhopping step 79: f -81.8418 trial_f -81.8418 accepted 1 lowest_f -81.8441\n", + "basinhopping step 80: f -81.8441 trial_f -81.8441 accepted 1 lowest_f -81.8441\n", + "basinhopping step 81: f -81.8441 trial_f -81.8326 accepted 0 lowest_f -81.8441\n", + "basinhopping step 82: f -81.8423 trial_f -81.8423 accepted 1 lowest_f -81.8441\n", + "basinhopping step 83: f -81.8199 trial_f -81.8199 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 84: f -81.8199 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 85: f -81.8199 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 86: f -81.8406 trial_f -81.8406 accepted 1 lowest_f -81.8441\n", + "basinhopping step 87: f -81.8366 trial_f -81.8366 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 88: f -81.8366 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 89: f -81.8366 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 90: f -81.8343 trial_f -81.8343 accepted 1 lowest_f -81.8441\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 91: f -81.8343 trial_f 1e+06 accepted 0 lowest_f -81.8441\n", + "basinhopping step 92: f -81.8362 trial_f -81.8362 accepted 1 lowest_f -81.8441\n", + "basinhopping step 93: f -81.8384 trial_f -81.8384 accepted 1 lowest_f -81.8441\n", + "basinhopping step 94: f -81.8444 trial_f -81.8444 accepted 1 lowest_f -81.8444\n", + "found new global minimum on step 94 with function value -81.8444\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 95: f -81.8444 trial_f 1e+06 accepted 0 lowest_f -81.8444\n", + "basinhopping step 96: f -81.8444 trial_f -51.7093 accepted 0 lowest_f -81.8444\n", + "basinhopping step 97: f -81.842 trial_f -81.842 accepted 1 lowest_f -81.8444\n", + "warning: basinhopping: local minimization failure\n", + "basinhopping step 98: f -81.842 trial_f 1e+06 accepted 0 lowest_f -81.8444\n", + "basinhopping step 99: f -81.8402 trial_f -81.8402 accepted 1 lowest_f -81.8444\n", + "adaptive stepsize: acceptance rate 0.450000 target 0.500000 new stepsize 0.081 old stepsize 0.09\n", + "basinhopping step 100: f -81.8428 trial_f -81.8428 accepted 1 lowest_f -81.8444\n" + ] + } + ], + "source": [ + "MAP_pool = sc.get_map_impalapool(setup, n_samples = 1000, theta_init = mat_trans[np.argmax(POST_pool),:].reshape(1,-1), niter = 100)" + ] + }, + { + "cell_type": "markdown", + "id": "590fe30b", + "metadata": {}, + "source": [ + "### Evaluate Results\n", + "Now, we evaluate the results. First, we compare the prediction error we get from the MAP estimation vs. the theta posterior median and the \"best\" posterior draw in terms of the joint IMPALA posterior. In this example, the posterior median provides a poorly-predictive value of theta due to the unusual structure of the posterior distribution for theta. The best posterior draw and the MAP estimate produce very similar prediction error, albeit with different values of poorly-identified theta. " + ] + }, + { + "cell_type": "code", + "execution_count": 43, + "id": "22163693", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + " 0 1 2 3 4 5 6 \\\n", + "0 0.090752 0.456076 0.600826 0.511963 0.498738 0.489947 0.507529 \n", + "1 0.090723 0.919064 0.365493 0.710719 0.577680 0.474112 0.710779 \n", + "2 0.090710 0.973705 0.190506 0.865703 0.895570 0.532328 0.786087 \n", + "\n", + " 7 8 method sse mape \n", + "0 0.504030 0.502311 parent_median 4.480262 17.494239 \n", + "1 0.794391 0.674195 parent_minsse 0.011477 0.696793 \n", + "2 0.679961 0.928478 map 0.011477 0.697251 \n" + ] + } + ], + "source": [ + "BEST = dict()\n", + "SSE = dict()\n", + "MAPE = dict()\n", + "n_exp = len(np.unique(setup.s2_ind))\n", + "s2_inds = setup.s2_ind[0]\n", + "THETA_Y = sc.get_outcome_predictions_impala(setup, theta_input = mat_trans)['outcome_draws']\n", + "\n", + "### Parent_Median\n", + "pred_median = sc.get_outcome_predictions_impala(setup, theta_input = np.median(mat_trans,axis = 0).reshape(1,-1))['outcome_draws']\n", + "median_sse = sum([((pred_median[0][0,np.where(s2_inds == i)]-setup.ys[0][np.where(s2_inds == i)])**2).mean(axis=1) for i in range(n_exp)])\n", + "median_mape = 100*np.mean([(np.abs(pred_median[0][0,np.where(s2_inds == i)]-setup.ys[0][np.where(s2_inds == i)])/setup.ys[0][np.where(s2_inds == i)]).mean(axis=1) for i in range(n_exp)])\n", + "BEST['parent_median'] = np.median(mat_trans,axis = 0)\n", + "SSE['parent_median'] = median_sse\n", + "MAPE['parent_median'] = median_mape\n", + "\n", + "### Best Draw \n", + "parent_sse = sum([((THETA_Y[0][:,np.where(s2_inds == i)[0]]-setup.ys[0][np.where(s2_inds == i)])**2).mean(axis=1) for i in range(n_exp)])\n", + "pred_minsse = np.hstack([THETA_Y[0][np.where(parent_sse == parent_sse.min()),np.where(s2_inds == i)[0]]for i in range(n_exp)]).reshape(1,-1)\n", + "parent_mape = 100*np.mean([(np.abs(pred_minsse[0,np.where(s2_inds == i)]-setup.ys[0][np.where(s2_inds == i)])/setup.ys[0][np.where(s2_inds == i)]).mean(axis=1) for i in range(n_exp)])\n", + "BEST['parent_minsse'] = mat_trans[np.where(parent_sse == parent_sse.min())[0],:].flatten()\n", + "SSE['parent_minsse'] = np.array([parent_sse.min()])\n", + "MAPE['parent_minsse'] = parent_mape\n", + "\n", + "### MAP\n", + "pred_map = sc.get_outcome_predictions_impala(setup, theta_input = np.array(pd.DataFrame(MAP_pool.values())).reshape(1,-1))['outcome_draws']\n", + "map_sse = sum([((pred_map[0][0,np.where(s2_inds == i)]-setup.ys[0][np.where(s2_inds == i)])**2).mean(axis=1) for i in range(n_exp)])\n", + "map_mape = 100*np.mean([(np.abs(pred_map[0][0,np.where(s2_inds == i)]-setup.ys[0][np.where(s2_inds == i)])/setup.ys[0][np.where(s2_inds == i)]).mean(axis=1) for i in range(n_exp)])\n", + "BEST['map'] = np.array(pd.DataFrame(MAP_pool.values())).flatten()\n", + "SSE['map'] = map_sse\n", + "MAPE['map'] = map_mape\n", + "\n", + "BEST_df = pd.DataFrame(BEST.values(), columns = np.array(pd.DataFrame(setup.bounds.keys())).flatten())\n", + "BEST_df['method'] = np.array(pd.DataFrame(BEST.keys())).flatten()\n", + "BEST_df['sse'] = np.array(pd.DataFrame(SSE.values())).flatten()\n", + "BEST_df['mape'] = np.array(pd.DataFrame(MAPE.values())).flatten()\n", + "print(BEST_df)\n" + ] + }, + { + "cell_type": "markdown", + "id": "fb8b0a6d", + "metadata": {}, + "source": [ + "## Exploring the Identifiability of the Posterior Mode\n", + "As mentioned earlier, the proposed MAP estimation algorithm is not guaranteed to identify a unique posterior mode; indeed, many calibration problems do not have a unique posterior mode. The degree to which the identified posterior mode can be viewed as \"unique\" can be roughly explored using various posterior visualization and summarization diagnostics. \n", + "\n", + "First, we visualize the integrated posterior distribution of the first three elements of theta. Some unidentifiability or weak identifiability is evident, as the posterior distribution has a ridge along the maximum rather than a unique maximizer." + ] + }, + { + "cell_type": "code", + "execution_count": 55, + "id": "f1bfff21", + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "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", + "text/plain": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "with open('/Users/lbeesley/Desktop/tame_impala/Helper_Functions/Posterior_Mode.py') as fd:\n", + " exec(fd.read()) \n", + "\n", + "\n", + "theta_grid = np.array([(np.linspace(0,1,50)[x], np.linspace(0,1,50)[y], np.linspace(0,1,50)[z], 0.5, 0.5, 0.5, 0.5, 0.5, 0.5) for x in range(50) for y in range(50) for z in range(50)])\n", + "GRID_pool = sc.eval_partialintlogposterior_impalapool(setup, n_samples = 1000, theta = theta_grid)\n", + "\n", + "from mpl_toolkits.mplot3d import Axes3D\n", + "from matplotlib import cm\n", + "from matplotlib.ticker import LinearLocator\n", + "X = theta_grid[:,0]\n", + "Y = theta_grid[:,1]\n", + "Z = theta_grid[:,2]\n", + "X0 = X[ ~np.isnan(X+Y+Z) ]\n", + "Y0 = Y[ ~np.isnan(X+Y+Z) ]\n", + "Z0 = Z[ ~np.isnan(X+Y+Z) ]\n", + "P0 = GRID_pool.flatten()[ ~np.isnan(X+Y+Z) ]\n", + "fig, ax = plt.subplots(subplot_kw={\"projection\": \"3d\"}, figsize=(15,5))\n", + "surf = ax.scatter3D(X0,Y0,P0,c=Z0,cmap=cm.coolwarm)\n", + "fig.colorbar( surf, shrink=0.5, aspect=5, label = 'THETA3')\n", + "ax.set_xlim(np.nanmin(X0), np.nanmax(X0))\n", + "ax.set_ylim(np.nanmin(Y0), np.nanmax(Y0))\n", + "ax.set_zlim(np.nanmin(P0), np.nanmax(P0))\n", + "ax.set_xlabel('THETA1')\n", + "ax.set_ylabel('THETA2')\n", + "ax.set_zlabel('log(integrated posterior)')\n", + "ax.set_title('Visualization of Posterior Surface')\n", + "plt.show()\n" + ] + }, + { + "cell_type": "markdown", + "id": "4d165122", + "metadata": {}, + "source": [ + "Next, we summarize the theta values associated with quantiles of the posterior surface. Variability approaching zero as the integrated posterior quantiles approach 1 indicates good identifiability of the posterior mode. This evaluation can be implemented either using IMPALA posterior draws for theta or across a Latin Hypercube sampled grid of theta values. The magnitude of the variances obtained by the two methods cannot be very easily compared to each other. In both cases, the absolute value of the variances can be difficult to directly interpret; rather, the relative orderings of the parameter variances as the posterior quantiles approach 1 can provide a sense of the relative identifiability of the posterior modes for each parameter. The horizontal dashed line indicates the uncertainty we would see if the posterior distribution was a Uniform(0,1) prior, indicating a total lack of information about a given parameter.\n", + "\n", + "Here, we use the posterior samples of theta to evaluate the identifiability of the posterior mode. The posterior mode for theta1 appears well-identified, and theta3 appears to be slightly better identified than some other parameters. All other parameters exhibit very worrisome behavior indicating at best weak identifiability. " + ] + }, + { + "cell_type": "code", + "execution_count": 56, + "id": "4e05d5c6", + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAiEAAAGHCAYAAABmuoLpAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8pXeV/AAAACXBIWXMAAA9hAAAPYQGoP6dpAADGvUlEQVR4nOydd5wURfr/3909OezO5l0WWGBBMhJEFAMiEswZ9YLxTv0aznym0zNjulN/3uF5nid63inGUwyAIqIIEgRUgpLj5jg7OXT9/pid2TRpA0Gv36/Xvnamu7qquqa769PP81SVJIQQaGhoaGhoaGgcYOSDXQENDQ0NDQ2N/000EaKhoaGhoaFxUNBEiIaGhoaGhsZBQRMhGhoaGhoaGgcFTYRoaGhoaGhoHBQ0EaKhoaGhoaFxUNBEiIaGhoaGhsZBQRMhGhoaGhoaGgcFTYRoaGhoaGhoHBQ0EdIDrFixgrPPPpu+fftiNBopKCjg6KOP5pZbbmmTbvbs2cyZM+fgVLKZ++67D0mSDkhZO3fuRJKklOf8+eefI0lS7M9gMJCXl8cxxxzD3Xffza5du/Z7XaN1+Pzzzzt97MaNG7nvvvvYuXNnh32XXnop/fr163b99id1dXVceOGF5OfnI0kSZ511Voc0c+bMafMbJfqLnuull16KzWbr0Xoe7Psneu/U1NQctDp0hfb3l6IoFBQUcP7557Np06b9UuYjjzzCf//73/2Sd7rPlZ5kz549XHfddZSWlmIymcjKyuLEE09k7ty5B6wO6ZKo7eM94w5kf5AI3UEt/WfAhx9+yBlnnMEJJ5zA448/TlFREeXl5axevZrXX3+dP/3pT7G0s2fPJjc3l0svvfTgVfgQ5pFHHmHy5MmEw2Fqa2tZsWIF//znP3nqqad44YUX+OUvf7nfyh47dizLly9n2LBhnT5248aN3H///ZxwwgkdBMc999zDDTfc0EO13D88+OCDvPvuu/zzn/+ktLSU7OzsDmlOPfVUli9f3mbb0UcfzXnnnddGbBuNxv1WT+3+6R7R+ysQCLB69WoeeOABFi1axPfff09xcXGPl3XeeefFFbTdpaioiOXLl1NaWtrjecfjq6++4rTTTsNms3HbbbcxatQoGhsbeeONN7jwwgv56KOPYiL9UCBR23fnGbc/0URIN3n88cfp378/CxYsQKdrac4LL7yQxx9//CDW7MDg8XiwWCw9ktegQYM46qijYt/POOMMbrnlFk466SQuvfRSRo0axciRI3ukrPZkZGS0KbunOFAPyu6wfv16SktLk4q8vLw88vLyOmwvKCjYL+2m0fO0vr+OP/54HA4HV1xxBXPmzOHuu+8+yLVLTTgcJhQKYTQae/Sa83q9mEymuCKioaGBc845h8zMTFasWEFBQUFs35lnnsmoUaO44447GD16NDfddFOP1Wl/sL+ecd1Fc8d0k9raWnJzc9sIkCiy3NK8/fr1Y8OGDSxZsqSD6drn83HLLbcwevRoMjMzyc7O5uijj+a9997rkKckSVx33XX861//YujQoVgsFg4//HA++OCDDmk//PBDRo8ejdFopH///jz55JNxz+Gvf/0rxx9/PPn5+VitVkaOHMnjjz9OMBhsk+6EE05gxIgRfPHFF0ycOBGLxcLll18OQFlZGTNnzsRut5OZmckFF1xARUVF2u2YiOzsbJ5//nlCoRBPPfVUm31btmzhF7/4Bfn5+RiNRoYOHcpf//rX2P7q6moMBgP33HNPh3x/+OEHJEni//2//wfEN1WuXr2aCy+8kH79+mE2m+nXrx8XXXRRG/fQnDlzOP/88wGYPHly7LeNmorjuWN8Ph933nkn/fv3x2AwUFxczLXXXktDQ0ObdP369eO0005j/vz5jB07FrPZzJAhQ/jnP/+ZVtvV1dVxzTXXUFxcjMFgYMCAAdx99934/X6gxaz96aefsmnTpljdu+KSSsTWrVs55ZRTsNls9OnTh1tuuSVWfpRAIMBDDz3EkCFDMBqN5OXlcdlll1FdXd2mLXri/onHJ598wplnnknv3r0xmUwMHDiQq666KqHbZc+ePZxzzjlkZGSQmZnJr371qzZ1BVBVlccffzx2Tvn5+Vx88cXs3bs3lubGG2/EarXidDo7lHHBBRdQUFDQ5h6cO3cuRx99NFarFZvNxvTp01m7dm1a5xiPaIcUvZ7TqTPA2rVrOe2002L3Xa9evTj11FNj6SRJwu128/LLL8d+qxNOOCF2fEVFBVdddRW9e/fGYDDQv39/7r//fkKhUCxN9Np8/PHHeeihh+jfvz9Go5HFixcndMcsXbqUKVOmYLfbsVgsTJw4kQ8//LBNmqjFYuHChVx++eXk5eVhsVg6XJNR/vGPf1BVVcWjjz7aRoBE+f3vf8+QIUOYNWtWrP7RMtq7Z+M9Y9K99qJukw0bNnDRRReRmZlJQUEBl19+OY2NjbF0ydq+My7nnr7WkiI0usVvfvMbAYjrr79efP311yIQCMRNt2bNGjFgwAAxZswYsXz5crF8+XKxZs0aIYQQDQ0N4tJLLxX/+te/xGeffSbmz58vbr31ViHLsnj55Zfb5AOIfv36iSOPPFK88cYb4qOPPhInnHCC0Ol0Ytu2bbF0n376qVAURRx77LHinXfeEW+++aYYP3686Nu3r2j/s990003iueeeE/PnzxefffaZeOqpp0Rubq647LLL2qSbNGmSyM7OFn369BHPPvusWLx4sViyZInweDxi6NChIjMzUzz77LNiwYIF4ne/+12srJdeeilpGy5evFgA4s0330yYpqioSJSWlsa+b9iwQWRmZoqRI0eKV155RSxcuFDccsstQpZlcd9998XSnX322aJPnz4iHA63ye/3v/+9MBgMoqampk0dFi9eHEvz5ptvinvvvVe8++67YsmSJeL1118XkyZNEnl5eaK6uloIIURVVZV45JFHBCD++te/xn7bqqoqIYQQl1xyiSgpKYnlqaqqmD59utDpdOKee+4RCxcuFE8++aSwWq1izJgxwufzxdKWlJSI3r17i2HDholXXnlFLFiwQJx//vkCEEuWLEnapl6vV4waNUpYrVbx5JNPioULF4p77rlH6HQ6ccoppwghhPD5fGL58uVizJgxYsCAAbG6NzY2Js07CiCuvfbauPsuueQSYTAYxNChQ8WTTz4pPv30U3HvvfcKSZLE/fffH0sXDofFjBkzhNVqFffff7/45JNPxD/+8Q9RXFwshg0bJjwejxCi5+6feDz33HNi1qxZ4v333xdLliwRL7/8sjj88MPF4MGD29zPf/zjHwUgSkpKxG233SYWLFgg/vznP8d+u9Zpr7zySgGI6667TsyfP1/87W9/E3l5eaJPnz6xa+fbb78VgHjhhRfa1Ke+vl4YjUZx8803x7Y9/PDDQpIkcfnll4sPPvhAvPPOO+Loo48WVqtVbNiwIen5Jbq/3nvvPQGIu+66K+06u1wukZOTI4444gjxxhtviCVLloi5c+eKq6++WmzcuFEIIcTy5cuF2WwWp5xySuy3itaxvLxc9OnTR5SUlIjnn39efPrpp+LBBx8URqNRXHrppbG67dixQwCiuLhYTJ48Wbz11lti4cKFYseOHbF9rZ8rn3/+udDr9WLcuHFi7ty54r///a+YNm2akCRJvP7667F0L730UizfK6+8Unz88cfirbfeEqFQKG7bTZs2TSiKIlwuV8L2/f3vfy8AsXLlyjZl7NixI+7v0PoZ09lrb/DgweLee+8Vn3zyifjzn/8sjEZjm+d0sraPV34039Z051rrCpoI6SY1NTXi2GOPFYAAhF6vFxMnThSzZs0STU1NbdIOHz5cTJo0KWWeoVBIBINBccUVV4gxY8a02QeIgoIC4XQ6Y9sqKiqELMti1qxZsW0TJkwQvXr1El6vN7bN6XSK7OzsDhdda8LhsAgGg+KVV14RiqKIurq62L5JkyYJQCxatKjNMc8995wAxHvvvddm+29/+9seEyETJkwQZrM59n369Omid+/eHTrM6667TphMpli933//fQGIhQsXxtKEQiHRq1cvce6553aoQ+sbtD2hUEi4XC5htVrFM888E9v+5ptvJjy2vQiZP3++AMTjjz/eJt3cuXMFIP7+97/HtpWUlAiTySR27doV2+b1ekV2dra46qqrEtZTCCH+9re/CUC88cYbbbY/9thjHdpj0qRJYvjw4Unzi0cqERKv/FNOOUUMHjw49v21114TgHj77bfbpFu1apUAxOzZs2PbeuL+SYWqqiIYDIpdu3Z1uKajD+ybbrqpzTH//ve/BSBeffVVIYQQmzZtEoC45ppr2qRbsWJFm05fCCHGjh0rJk6c2Cbd7NmzBSC+//57IYQQu3fvFjqdTlx//fVt0jU1NYnCwkIxc+bMpOcUvbbnzp0rgsGg8Hg84osvvhADBw4UiqKIb7/9Nu06r169WgDiv//9b9IyrVaruOSSSzpsv+qqq4TNZmtzTQshxJNPPimAWCcXFRqlpaUdXuziiZCjjjpK5Ofnt3nmhkIhMWLECNG7d2+hqqoQokUgXHzxxUnrH2XIkCGisLAwaZro8y/6/OqMCGlNOtde++fGNddcI0wmU+z8hEjc9umIkO5ea11Bc8d0k5ycHL788ktWrVrFo48+yplnnsnmzZu58847GTlyZNqR9G+++SbHHHMMNpsNnU6HXq/nxRdfjBu9PnnyZOx2e+x7QUEB+fn5MbOq2+1m1apVnHPOOZhMplg6u93O6aef3iG/tWvXcsYZZ5CTk4OiKOj1ei6++GLC4TCbN29ukzYaFd6axYsXY7fbOeOMM9ps/8UvfpHWuaeDECL22efzsWjRIs4++2wsFguhUCj2d8opp+Dz+fj6668BOPnkkyksLOSll16KHb9gwQLKyspirqREuFwubr/9dgYOHIhOp0On02Gz2XC73V0eVfDZZ58BdAiuPP/887FarSxatKjN9tGjR9O3b9/Yd5PJxGGHHZZyxNBnn32G1WrlvPPOa7M9Wm77cvYHkiR1uN5GjRrVpu4ffPABDoeD008/vc3vOHr0aAoLC9N2DXXm/mlPVVUVV199NX369IkdW1JSAhD3+PaxMzNnzkSn07F48WKA2P/2v/GRRx7J0KFD27T9ZZddxrJly/jxxx9j21566SXGjx/PiBEjgMj1GgqFuPjii9u0kclkYtKkSWm30QUXXIBer8disXD88ccTDod56623GDVqVNp1HjhwIFlZWdx+++387W9/Y+PGjWmVHeWDDz5g8uTJ9OrVq825nHzyyQAsWbKkTfozzjgDvV6fNE+3282KFSs477zz2ozIUhSFX//61+zdu7dN+wKce+65nap3MqLPpq4Epnb22mv/jB01ahQ+n4+qqqou1LwjPXWtdQYtMLWHOOKIIzjiiCMACAaD3H777Tz11FM8/vjjKQNU33nnHWbOnMn555/PbbfdRmFhITqdjueeey6u/z8nJ6fDNqPRiNfrBaC+vh5VVSksLOyQrv223bt3c9xxxzF48GCeeeYZ+vXrh8lkYuXKlVx77bWxPKMUFRV1yLO2tjauvzRe+V1l9+7d9OrVK1ZeKBTi2Wef5dlnn42bPir+dDodv/71r3n22WdpaGjA4XAwZ84cioqKmD59etIyf/GLX7Bo0SLuuecexo8fT0ZGBpIkccopp3Rol3Spra1Fp9N1CPKUJInCwkJqa2vbbE/1Wycrp7CwsMODMT8/H51O16Gc/YHFYmkjgiFSd5/PF/teWVlJQ0MDBoMhbh7piPjO3j+tUVWVadOmUVZWxj333MPIkSOxWq2oqspRRx0Vt53bX9c6nY6cnJxYm0b/x7tXevXq1UaE/fKXv+TWW29lzpw5zJo1i40bN7Jq1Spmz54dS1NZWQnA+PHj455D69izZDz22GOceOKJKIpCbm4uffr0ie1Lt86ZmZksWbKEhx9+mLvuuov6+nqKior47W9/yx/+8IeUgqGyspJ58+YlTNf+945Xn/bU19cjhEhY99bn15l8Afr27cuWLVtwu91Yrda4aaKxH63bMx26cu21fx5ER6N19XnUnp661jqDJkL2A3q9nj/+8Y889dRTrF+/PmX6V199lf79+zN37tw2nUaiYKlUZGVlIUlS3MDQ9tv++9//4na7eeedd2IKHGDdunVx846n9nNycli5cmXKsrrKypUrqaio4IorrgAi5xd9y7n22mvjHtO/f//Y58suu4wnnniC119/nQsuuID333+fG2+8EUVREpbZ2NjIBx98wB//+EfuuOOO2Ha/309dXV2XzyUnJ4dQKER1dXUbISKEoKKiIuHN35VyVqxYgRCizW9WVVVFKBQiNze3R8rpLrm5ueTk5DB//vy4+1tb/BLRnftn/fr1fPvtt8yZM4dLLrkktn3r1q0Jj6moqGgzpDUUClFbWxvrIKL/y8vL6d27d5tjy8rK2rR9VlYWZ555Jq+88goPPfQQL730EiaTiYsuuiiWJpr+rbfeanOPdpYBAwbEXpTa05k6jxw5ktdffx0hBN999x1z5szhgQcewGw2t7lX4pGbm8uoUaN4+OGH4+6PioYo6VgXsrKykGWZ8vLyDvvKyspi5XY2X4Bp06axcOFC5s2bx4UXXthhvxCC999/n5ycHA4//HCAmPBuf/21F1hdufb2Nz11rXUGzR3TTeJd+NBiSmt9UyV6g41O0NX6xqioqEg7ur89VquVI488knfeeafNW2dTUxPz5s3rUHa0blGEELzwwgtplzd58mSampp4//3322z/z3/+05Xqt6Guro6rr74avV4fGwJnsViYPHkya9euZdSoUTErVOu/1m8MQ4cOZcKECbz00kv85z//we/3c9lllyUtV5IkhBAd5r34xz/+QTgcbrOtM28jU6ZMASIdZ2vefvtt3G53bH93mTJlCi6Xq8OkRa+88kqbehxsTjvtNGprawmHw3F/x8GDB8fS7o/7J971D/D8888nPObf//53m+9vvPEGoVAoNgoh6q5s/xuvWrWKTZs2dWj7yy67jLKyMj766CNeffVVzj77bBwOR2z/9OnT0el0bNu2LW4bJRIWnaGzdYZI2x1++OE89dRTOBwO1qxZE9uX6Lc67bTTYkPC451HexGSDlarlQkTJvDOO++0KVNVVV599VV69+7NYYcd1ul8Aa644goKCgq4884747o8Hn/8cX744Qeuvvrq2DUUHbX13XfftUnb/vnYlWsvHdKxlCbiQFxr7dEsId1k+vTp9O7dm9NPP50hQ4agqirr1q3jT3/6Ezabrc1EVdE3iLlz5zJgwABMJhMjR47ktNNO45133uGaa67hvPPOY8+ePTz44IMUFRWxZcuWLtXrwQcfZMaMGUydOpVbbrmFcDjMY489htVqbfMmP3XqVAwGAxdddBG///3v8fl8PPfcc9TX16dd1sUXX8xTTz3FxRdfzMMPP8ygQYP46KOPWLBgQafqvGXLFr7++mtUVY1NVvbiiy/idDp55ZVXGD58eCztM888w7HHHstxxx3H//3f/9GvXz+amprYunUr8+bNi8VeRLn88su56qqrKCsrY+LEiW06t3hkZGRw/PHH88QTT5Cbm0u/fv1YsmQJL774YpsOAoj57v/+979jt9sxmUz0798/ritl6tSpTJ8+ndtvvx2n08kxxxzDd999xx//+EfGjBnDr3/96061WSIuvvhi/vrXv3LJJZewc+dORo4cydKlS3nkkUc45ZRTOOmkk3qknO5y4YUX8u9//5tTTjmFG264gSOPPBK9Xs/evXtZvHgxZ555JmeffTawf+6fIUOGUFpayh133IEQguzsbObNm8cnn3yS8Jh33nkHnU7H1KlT2bBhA/fccw+HH344M2fOBGDw4MFceeWVPPvss8iyzMknn8zOnTu555576NOnT4f5JKZNm0bv3r255pprqKio6CCQ+/XrxwMPPMDdd9/N9u3bmTFjBllZWVRWVrJy5UqsViv3339/V5o/Rrp1/uCDD5g9ezZnnXUWAwYMQAjBO++8Q0NDA1OnTo3lN3LkSD7//HPmzZtHUVERdrudwYMH88ADD/DJJ58wceJEfve73zF48GB8Ph87d+7ko48+4m9/+1sHS0w6zJo1i6lTpzJ58mRuvfVWDAYDs2fPZv369bz22mtdnkjM4XDw9ttvc9pppzFu3Dhuu+02Dj/8cJxOJ3PnzuXf//43U6dO5b777osdM378eAYPHsytt95KKBQiKyuLd999l6VLl7bJuyvXXjokavt0OBDXWgd6PNT1f4y5c+eKX/ziF2LQoEHCZrMJvV4v+vbtK37961/HhqxF2blzp5g2bZqw2+2xoX5RHn30UdGvXz9hNBrF0KFDxQsvvBB3+BQJRiSUlJR0iIh+//33xahRo4TBYBB9+/YVjz76aNw8582bJw4//HBhMplEcXGxuO2228THH3/cIZI62SiKvXv3inPPPVfYbDZht9vFueeeK5YtW9ap0THRP51OJ3JycsTRRx8t7rrrLrFz5864x+3YsUNcfvnlori4WOj1epGXlycmTpwoHnrooQ5pGxsbhdlsjjsksnUdWp9v9JyysrKE3W4XM2bMEOvXr4/b1k8//bTo37+/UBSlzTm3Hx0jRGSEy+233y5KSkqEXq8XRUVF4v/+7/9EfX19m3QlJSXi1FNP7VDXSZMmpTVKpLa2Vlx99dWiqKhI6HQ6UVJSIu688842w4Cj+e2P0TFWq7XD9njXXzAYFE8++WTsGrTZbGLIkCHiqquuElu2bIml64n7Jx4bN24UU6dOFXa7XWRlZYnzzz9f7N69WwDij3/8Y4e6f/PNN+L000+PXesXXXSRqKysbJNnOBwWjz32mDjssMOEXq8Xubm54le/+pXYs2dP3DrcddddAog7nDzKf//7XzF58mSRkZEhjEajKCkpEeedd5749NNPk55fOqPP0q3zDz/8IC666CJRWloqzGazyMzMFEceeaSYM2dOm7zWrVsnjjnmGGGxWATQ5nqtrq4Wv/vd70T//v2FXq8X2dnZYty4ceLuu++ODYWNjoB54oknOtQz3ugYIYT48ssvxYknniisVqswm83iqKOOEvPmzWuTJjpyZdWqVUnboj27du0S11xzTazO0WfVAw88EHd47+bNm8W0adNERkaGyMvLE9dff7348MMPOzxjOnvtRYdKtz+f1iNxErV9ukN0hej6tdYVJCFaDTvQ0NDQ0NDQSMr333/Pcccdx+jRo/n4448xm80Hu0o/WbSYEA0NDQ0NjU4wcuRI3nvvPb7++mvOOeccAoHAwa7STxbNEqKhoaGhoaFxUNAsIRoaGhoaGhoHBU2EaGhoaGhoaBwUNBGioaGhoaGhcVDQRIiGhoaGhobGQUGbrCwOqqpSVlaG3W7v8iQ3GhoaGhoa/4sIIWhqaqJXr14p15vRREgcysrKOr0YkYaGhoaGhkYLe/bsSTkDriZC4hBdNGvPnj1kZGQc5NpoaGhoaGj0ANuXIN68DEkNEEDHn7gKi7UORQ4yYuTiSBohMCoOjjp2MZKUeJHPZDidTvr06ZPWApSaCIlD1AWTkZGhiRANDQ0NjZ8mahh2LQNXJbiqYNEDoA8CEuuz8rE2CY46+lOEkNu5TZpQ1R/JyjqqW8WnE86giRANDQ0NDY2fGxvfh/m3g7Osw66qHAMb8gpxKJVIEkiS2iGN399x1eD9gSZCNDQ0NDQ0fk5sfB/euJjIOnttCegk1g+1492bgSOro0CJYjTm78cKtqAN0dXQ0NDQ0Pi5oIYjFpA4AgRAHxIYQhJebwYORwUAbRdvkTAai3A4xu/3qoImQjQ0NDQ0NH4+7FoW1wUTRQIO2+IkHFYwm10ACCG12guHDbqny0GpnUUTIRoaGhoaGj8XXJUpk+TXBrCYg7HvshwxhRiNhYwc8Vfy86fvt+q1R4sJ0dDQ0NDQ+LlgzkqZxI8ei6089n1A/5twOI7A4Rh/wCwgUTQRoqGhoaGh8XNAVWHNKykSSdRaB7WKBzHTr9//HXDxEUVzx2hoaGhoaPwc+PRe2PjfJAkiMR9Vh5+BXh8AQJaHHzQBApoIOTCoYdjxJXz/VuS/Gv5pl6OhoaGhcWix4u+w7NnIZ0mGY2+GjF5t02T0gpmvUG9oiG2y2w7MKJhEaO6Y/U28CWMyesGMx2DYGS3bWs9sZyuAkokgt1OnydL0ZDkHgkOlHhoaGho/RVo/Q+u2w+JHWvad+ic44nI48Q9xn7OBvbPRNff++fmTDk79m9FEyP4k0YQxzvLI9pmvRARCOgIiWRrouXLS4UAJJg2NdNAErcb/GklmQ+XYmyMCBCL3Qf/j2uwOh30oynYAfD4rBQUj93dtk6KJkP1F0gljmrd9eDO4a+DDmzomaS0gIInI+DWYs5OX89GtEPTCu1cmLycqALorIHpCMKWqh4YGaIJW43+PJLOhAlA0OunhDY2rkaSIq97l6ovJZOrZ+nUSSQiR4Ez+d3E6nWRmZtLY2Nj1Bex2fAkvn9b9yuitIEkQcHU/r2RkFMON38MPH3ZeQDQHOyUVTEiRbeZs8NYlqIQUKStVPTrjXtKEzM+XhA/jVtejJmg1fk6oYXh6RNLJyGLP8gTX9w8/PMy+sn8CUF11Nhde+GSPV7MzfahmCdlfpDFhTFoE3T2TTyqc++AfJ0HZmjj7yiIWF5ODpBaXD24GWUqeJqEAaU7j3AefPwpfPNExn866l35qb8nddXN1Js1PhUTnko6lcf4dMOTU9ATtz6nNNH6+pJgNFYg8Q3ct6+CGiVJbuzT22Wwe3YOV6xqaCNlf2AoOdg06TzwB0hpfQ/L9nuqeqccXjyfYIQAp0rkIFd68lIRCZeL1zZHiabh9DgW66+bqTJpDia64/qY/Cp6a9B7G/z4fti2Ks6+du/NAtdn/mojU6FnSfblNkC4QqMPn3xxJ0pRNdnZJT9Wsy2jumDj0iDsmZjYrJ/7bmgSWnMjDdL9yoMo5wMh6UINJEiSyyDTvi7p9om/V3X3od6dzScetAD2TZtgZPdfJddcV1pXYoZ7G5EggrveDS+fnKCI1Dizpuvkv+SCuJaSy8kPWb/gdAHv2DGP8EY8yfPjwnq6l5o45JJCVyIPjjYvp2CE2P+BO+RMsvDO5ULEXRZInS2POAm998/culmPKTG3pOBAoBggHUqdLKkAgeeclWkyW3vruP/S707lMmxX5bZK5Fd65MhIXlCzNu1c1f06UppUFacGd3Y+16a4rLOHIsWbXn8Ge4Fx6mITXfKs2S9elk4x0RspB+kHbPyUOFcvOgXrZ2J+UTARrHrgTWZ2bX7BKJsbdW1fX4oppqC8iOzt7P1Syc2iWkDj0iCUkStyHcTHMeLTdwxjiCog2D6cUabpTzgl3wuetxpl3mh4STN2uRyew5ICnNs6OTrwFd8uKcajQ7ny7KiCi+SRyhUX3nzcHFtwBTeV0C4MtecC2MQP8zu6VATDiPFj/Vpwdca6TeKQTTKgzR7ILehMkaGe9O1TorlhNJ4+eoCcsTIfKXEwvTIF9q+PsSH49CiFYtnwSPt8+VFVm+bILuP32ezAajT1Xt2Y604dqIiQOPSpCoIs3aisBkW6a7pQz5NTU7qNUAqInBFNa9cgGbzzx0MOkGjEUtWIk61xMjogVI9ZuhyrNndz0R+LH2kSZ/hh8+UQK914yVxgg60ANdb2qUY66Br5+rvnLwRK0abj2di3rmZFykNDMflDorlg9UPE4nRlF1d089rc7bd8aeGFyc9FyxLIZK6ddf9AOj2cXy78+EYCG+kJ27DibW2+9tft1ioMmQrpJj4uQdDhQAWupAgG7a3GJ5tMdwZSqHufNSeFeovkGTeSeIP2OsHRK/MDGnyvtH2yHMpd8kMCdlq6gPQB1sRdBr7Hw44c9U865L8LI83omr+6QqlOO3aOJBHrrl5puiINUpLRCpWFhSseSZc2H6Q9HXKf783z+dU7L8+jkxyF/WNr9wd59/+HHH+8BYMeOMSjyqVx22WXdq08CNBHSTQ6KCDlU6AmLS7ppulOPVEIl5hJIsP+o/4OvZ6dfn4NGmm6ulGl+TnQisDjpdRKdtyZeR9iMwQqBNIbJDz0DNr3f9VNKl4vnQb9jDm5cQjqdcrqxXYAKVJJHPRmY8VPCvsiiZj3hfupmIGen8khKD7jTdn4Fc06JfHb0Rb12Jfs2b8HVUI/NkUXx0OHISfL+7vtrqa6eD8DaNacwaNBJnHnmmV2rSwq0wFSNrjPsjMgbZLKHXJypgDuQTpru1GPYGZE3i7imz2ah0nt84v3mrENQhCQIYD659WiRrqY5xISJJbc5HqeLrr8Zj7ZcC8mutVTXCSRv14k3pOfS6ZYA6YSI/PwRqN8BTRUt2w50XEI6c1WkKUAgsopqEdUUUs06hrGQ4ziOVQxzbo3Nd6GqKrt27cLlcmGz2SgpKUGW01h/tZtDWjuVR1JaBcN35bkoBHz2YOxrRZ/zeO93/4errsUtasvO5cRLr2TQhI5BqUKEqa9fDkAwaMDlyiI7OxtVDbNv04a0hcz+QBMhGh3proA4UPVIR6gk2q+GIw/vZA/9bgc2pmnFiMZixB210soClUp0JUsz7ZHUI6TSHcqdVECQwhXWPvYkQed/+jOR/6nONx1SXSfJ2nXIqbBmTg9amLoiIluxe3nHbftrjahE9NREjO2QgDFsZDhb+JLxSKgMdVWyceNG5s//GKezKZY2I8POjBknM2zYsOSZWvPSK9yUGfkfT7zFDV5PTPO4qthnQavl6rvadlsXxX77gLUP/3ljJSJWSnPWdTW8/+dHOOPmuzoIkaamDYRCjQA0NBQCMoG6Gl649oq0hcz+QnPHxOF/2h3zv0SPjhhK0LmkG6yb7vwdPTIfSYJ6pIy1aS8gEuSTyhWWtLPsguuvJ+iOS2fQdNiyIHUZR10DG//btXiqoWfCiudILIRSBRZ3IS4hXpsgwcI/wNd/TX28Oad5huSOdVaBOZyPgSDZNJJDPSP4ASu+WJpGbDiH/ooXN0XXNmnd6Ua6+pkzZyYWIqoK834Ha/+Vuq4Fw2Hc5bD0T23b3pgRcceJcJKDWwT8RgaykOMYywaOZyUA9WRQQS5D2d61wGIh4O+ToPxbAD5pOJLvyhOPaLHn5PKbv7zYxqKxc+dzbNsemZ59y5YJVJQfhmX7BhR//NFY8YRMZ9BiQrqJJkL+h+juiKGEVowuBOseCHpiyHg6AiLd8z3Y8y6kS7LzMWelH3cQHS3TWRGZ7ggbvRWCblRgD0U4sWPD0/k4i3jna82LWAxqt6aoRGqxuoNiXub8dscJDATJo5YsGsmiEQdOguhYwwiqaG/VEGSYDdx4251IIkTDj3Pwu3djtPbFcdglSPPvhNUvpqhrd2kR8Bs//AtveI8BQEblct6gNxG32WpGYjGbGXbb/M5f3xvfi92TAccgnl1eAO2sIO2Zee8j9Bk+KvZ9zdpfxdwxq1aehc9nx/bDGqQEQejxhExn0ERIN9FEyP8Y3R0x1JNWjANBTwwZTyefQ+V8e4pk69ikI1a7E5T4/Vvw9hWdPqwJK4s4hu30YQZLGMbW2Nt4wjiLVKu0JiU9sfr9sN/z9tc7O5mziowgTNs2PO94Pw2BN/AbmjcIwdAtbnpVNFtVJBkm/B9sfLfjNT3xelj1ItRuSV64wQan/z/45A9x7wt1yGk8/cQsnN5ArA2yqedqXsVAZBTeu4ZzOPOOF9OLZYlea03lsOgBaNwDwO7R9/Lma4tTHn7K725j6DGTAAiHvSz5YixCBPD77axccRZSMIBt63dJ82gvZDqDFpiqodEZuhPYGO2UD0Swbk/R3VibdPM5VM63p0h0PunMjtw6kLYrdHEtKjtuzmIhO+jNPKYAMCwWZzEfp7Ml5ikjI4MZ06cxbMHtgEAFqskmiyYMtJqhWNbBL9+OxEului+GnYF62MnsWvUxrvoqbFn5lIw/mcC33wE7O3UuApmoU0RHCBtudJmNVIgloG9uaSEYtL1FgAhAOvt5GDUTpj0Y/5rOGgCvzUxeeMAFtny4cX3cPHbt2IHTG6S1haKOLD7kRM5mIQBTA/PZ88NaSoaNS15WPOEGkDMIaeBJQGoRYnNkxT43NKxGiEigcF1d5DqSA/6UebgaDszcRpoI0dBIRbqd8s+Jn5uA2N+kK1a7SsnElIHUqiGTBYHRGAiSSRMgMBHAjhsHTq7kNVYzkg1VId788o0OxzudTt548y1mYiaTfMz4KaAutj+ETB0O8tW6yPWRQGDIOn3smI5iZy/GzzcQCKQePWPUyQRCIQqoJgMXdTioIae5LjoayITGTHYum8wAxybsRfvIbfTRd1+LAPlhQAZDhp8VkQaJrulAU8dt8XBVJszD5Yo/c++3DGMYWxjMDmx4UL+8F4Z+0LwMQxySWaFqtxDe8VXKatpzcike2rIeTF19yzEN9UUAyAFfh+Pa01rI7E8OugiZPXs2TzzxBOXl5QwfPpynn36a446L//BbunQpt99+Oz/88AMej4eSkhKuuuoqbrrppjbp3n77be655x62bdtGaWkpDz/8MGefffaBOB2Nnytap6yRiv0pVltZW1RkdtELF1ZsuCmhDBnBrqMeZsUX25JmY8WN5cuvMGHFh6nNPj0BjuB7+lCBnbZzo2ynDx9yIrVkMZMPElhT9pKxfCszZsxg2LBhbNy4kTfe6Ch2/P7Ub+EApx8zjIVLllNBPo1kcCRrcdDETnqzgcGEmruvveHeOEN6BihBnP1rabQaGLm5kR8H2igvNJD/wxxyhl+ZuKB0rUxJ0tlstgR7JP7LNK7jFax4yShfCmtehnGXdkyqhiMiFoEAPCYZqy8SsyGAxoCZD1+fByS/niZfcmUslkOIMNXVC2N1iYyMAaOS3CXUXsjsTw6qCJk7dy433ngjs2fP5phjjuH555/n5JNPZuPGjfTt27dDeqvVynXXXceoUaOwWq0sXbqUq666CqvVypVXRi6y5cuXc8EFF/Dggw9y9tln8+677zJz5kyWLl3KhAkTDvQpamho/A+hIrGL3rhwYMNGCRJpRACkx7Az2DjxGeYv34BTWGKbMyQPM44eTp2hP5BchLix4sYad18QAxsZRCW55FBPDvWEUNhCf3bRm6iraT6TEDXw5pIE1pQ33uC8885jwYLkI4YMBgOnn346n3zySUe30IwZDBsyGHnl33jDewxeDKzLHIDB4EXyK/R37sSs81JjtpJVsoPs7BbrU22+ni8zcwk3DyApq9xMTrL+NKWVKfmicAAlJSUYDIa4Fh4vZt5jKr+geS6Z+XdByTGRuV7aBx87y6jO0uO1KPSqaBFr3xyWybpPBuILRcRFfv9S3A0NuOvbDh+WZJn8/gMAqKpawI+b7yMQqALA7zcSCkWEZ5Yjk4Z9iZuktZDZ3xzUwNQJEyYwduxYnnvuudi2oUOHctZZZzFr1qy08jjnnHOwWq3861+RYVgXXHABTqeTjz/+OJZmxowZZGVl8dprr6WVpxaYqqHRfbo8wVQrhAjT0LAKv78KozEfh2M8kqTslzTdPZ+EcRbNloHutkkiy0K66AlgIIQHE6Kb0kiSJHqi67jkkksoKSlJ3CYb3+ebRXdTNVDFYGxxIQQDBnRBkKyJ3TpCRLweDvFbxk25I3lF0g1AT8Dq1av54IMPmr+pDB7yFVu3TCAcNgAqDkc55ytVFNd+Hkmi6CHcKs4moxf0n4T3h7kYAipKqypUZBuYt2cYzt12ALJzM/nFE39HbzLFJhrbsWYVm5ZG8h5yzCTGXziK79df2+Zcamt7sXFDJC7IsmMDis8baaBWv6M9J5fJl3R/npCfRGBqIBDgm2++4Y472l4c06ZNY9myZWnlsXbtWpYtW8ZDDz0U27Z8+fIO7pnp06fz9NNPJ8zH7/e3MRG2foj83OiJh/GBQlUF5VsacDv9WDOMFA1yIMvJh6Z1hZ9Sm/xUSKdDTkVV1QI2b3kAv79ldlCjsZDDBt1Lfv70Hk3T3fNJJBCiloGZMyOBj+m0SbzrUQiJ+fPnp1XXCCoZGVUYjV4CATONjfkEMWAwmJkQWMkRfIsXC43YqcNBLVnUkEUVuQTRp8y9p95dXS4XkiRwOCoxmyPnK0ktVvCqXCMNwz0Y2hWn1wfA0HZbIGBg8+aJDBy4EpPJE+tfjQUlqSvSjZierVu38uGHLWsDlQ5cSX7+TnxeO7t2jQZk9PoAS+sCzLQVILkq2woQiJT57WuYW2+yKJQXmVi9p3dMgCiGEGdecj5GS8SaFR29MmDMEez8dg3eJic/fLUEpfc89I62jVZf3yv2WW7u7yb96nIK+pf+b86YWlNTQzgcpqCgrZ+toKCAioqKBEdF6N27N9XV1YRCIe677z5+85vfxPZVVFR0Os9Zs2Zx//33d+Eseo50OsLudso98TBOtx6p0qTav21tFV/O3YK7oUUcWh1GjrtgEKVj8tPOJxwOsW3DYtyuCqy2QkqHT0ZRWi77dNvkf02odOdaS6dDjna6idq1qmpBhzc5AL+/ku/XX8vIEZHJsnoiTX7+dMLhID/++AFudxlWay8GDz4NRdG3Ox+VzMwqDIZox67GXA8LFy4kGfPmzcPr9SbMI9omia5HR+ZVab0clZaWUl//OQMHrcRo9LQ6Xwvbto5n8sCJLFxxGD9Qyml8xgg2E0DHN4xkJ70J7pcuQWA0urFYGlFVhcbGfKJziErSWr5adnXc+y8v7yQ2b3kgsrH9pdf8vbUWMhgCDCxdzcYNkxh02Ers9lokCSqr/ohjn6C41wXJ7+EuxPRUVlby5ptvxkRZr+KNZJt3sXVeH4TiROfwEQqZqK7uR+HQTVRukilM0Vous8LGIXaa7JHfQqqOnrOg39R9SKX5HY4xWqwcde5FLJ7zPAC7vlQoPa1t/GttTUTcScEAklBx9MpjzIzTUXQHNzT0oAemSu2ihIUQHba158svv8TlcvH1119zxx13MHDgQC666KIu53nnnXdy8803x747nU769OnTmdNISqoOLJ2OMNIp/4jQr0cxNRD2OZCCIzjugsGxTjlZhxt5qF8TM1FG8fkq+H79NYwcMTtWVigUZu2yjdTXNZCV7WDMxGHodEqsHl/M3UyDpwpVDiCrBhyWfI6/4LBYPVIJiHT2z39+fYd2dDf4mf/8emZcNaJVPonb5LsVb1Be+zg6U2SoWZ0PdizIoijn94yaMDPtNknn90nWZocawUCItR+uwVnXSEZ2JmNOHYve0PIoSF8AdnQtAK3e2CNv43q9n1DIGOt85s+fz5AhQ6ip+SRuuw4a+Ae2bH2I+P75yLaNm+4EwinS/B5JkpOm2bzlQfbs2UVV9V8wGCLBmM4m2LX7j+Tn/44xoy9n/vz5ZGXtJSurHJcrB5/fihrWYTK5CAQsvPXWW4kbuxmv14vB4MaRVYHJ6MJkcpGbt4tQyMi2reOZP38+OTm7WL/h+g719fsrqay6n5ycSdTWdoyVa82QoV6czs9pb6gwGDwMG76E/PyRzFjxOW9wGq9yFnnU0xQLUo0cZMaLt807eVssFgsejyfh/tbpzOYfKB24Kq4gsljMVFa9Gud8I/ef3T6qzbURD0mCnTtHUVS0FaPRg8nsYsjQpaz/fgoDSr8hN3cPEObHH+9h69ZZhMMt9Yj3spEqpqf1Na8oCgsWLIhZ0XNydlNg2Uj56jwGzNiLrBcU1jWwu2wsdXW9KKsaSYXucwoTLN7t00t8NzwjIj5aPZByh9VTuTaHovHV2Ht78AWqO9TFZrMxcso01n78Pg2V5bjKrHhrjVhy/c1pJQKBiPUkOjJm3DkTDroAgYMYExIIBLBYLLz55pttRq7ccMMNrFu3jiVLlqSVz0MPPcS//vUvfvzxRwD69u3LTTfd1MYl89RTT/H000+za9eutPLsyZiQVB1Yoo4w+n3kiNk07RvD0g9eoWDM6+gtLWO3g54sKtdeyLGnXYw78DnltU+gM7UaUufLpijnNkYeeS5LPj+WkFoVd2SYEKCT85l0wlI+/2glS1d+hs2xJ/a25mrow7FHnkjfXqW8N+dzXBmbsWfvje1vquuNzXkYZ156AgDzn1+PkMIYitahmOsIe7MJlI9GEgqjp/Zh3Sd7ErbX9N8O58s3tuBpDAAqOpMTlAASEkF3DiBjdRg49vzD+OqjxG3S//AcAtaIm659uwLkWR+m3vtMyjYZOvwe1q+/Punvs2FVBktXfk5Yaum0FWHk2CNPYPKpR8e2pbLKpEM6YidZms9fWczKbV/jaVVXizByZOlRnHDx5JgAFAjMZieyHEBVDXi9GUhIMQEYcU98hCRtjV0HQgxk7Ngj+Pzzz7Faa9HrAzQ25iOEgtVaR1ZWGS5XFg0NxZx33gAqq+5P2K4Hkug10brccFjG6cyjseE0yspczb79nsVgcFPSbx35+dv5YdNxjBi5AVWtS5i+rGwQ27ZOoPXKJAaDm0DAQsSyoHLC5AWEw4nW/5EwGgs5ZkUNm5xW5jMJKdPT8vs1mpnBF2DOjs38GY+o5SdimWlv2YkIzYyMDCZNslPfEFkkMN49KEk2IP7Q1s7gyLyVJUtcSNK2WD1Cof4EgyGKi7+md5+NCdsDiFnD0nG5td8fJStTR3Hm29R852DAyXtQ9G27VZ/PSnnZYRzm3MYQ3R5MAZUMV2TWkwabjq0DLDhteoQu/sVfv82OY0ATkgS9sx4kbBwdt66H9+nFurdeBWDwedsx50Tuc48ng29WR1bM1ddXU2j5nnN+/zhZWUel0cKd5yczY+qECRMYN24cs2e3rGY6bNgwzjzzzLQDUx988EFefPFFdu7cCUQCU5uamvjoo49iaU4++WQcDscBD0xNJTBGDP8LGzfcT1hUJ+wIFSmHqu/OI2fk87Ht0bTRX865fQYZA+a32dd6f4bxIpy+1xGqAhJIUvOySpLatlzfXWzc9hWlpaswmlq9ufgsbNs2Hn/ZMRh7fUX/vt9BMJOwzw5IBAImKsoHIdePx6hmolrXkVWyEp3ejySHkZUAobBC7fZjCJQfEakbYfSF36GYGgg19SJUOwgJObL2WfyZhNsg6wKYsrdisFdisEWiv8NBM2rzX8hvQ4T1CCEjSSqKwYshYx8GeyV6ay06kxOdsQlJ50/R6ekQIpTw91HDDpZ9dSoAmY5WD+OGyMN40pHTmXzq0R2sMgAhX4tVBlKLlMUfLk8pdpKlkWp9fL6tWdy3X4YDmDTgeHZuNCCC5dSa9hEQAlUOEtZ5sAgjee5SZEMxR/wqi88+e7LNG67PZ6GyspSyfUMIhYztCmiLovjp03crubmbMJs7rl1xIIWIEBJ+v5lgwEwwZMDnteN05uH1ZqIoIfR6LwaDD73eh97gRa/3EQ4bcDbm4nLlEAiYkSQVv9+GTufHam1ApwsSCukJhowEAyaCQRPRkSVmsxO7vRazpRGj0R15ezc1YTR6Ep6zEFBZMYAdO45oblswGl0MGfolGRk1McuC0ZRBaem8lOc8NutagksfZfNAC35ji4A1+sMcttVD/pR/sJGBKTvkRYueTGjlOPHEm6hvuDqlJaMnGDvm32TYx7N1xUaa6p3YszIYOGEYVdVVzJkzh6zs7xk4cGXzc68jRmMRWY7nePPNt0kkqiZOnJg0VnF0iZ69Kxcz6PTd6EwRcREOSh3EiKpKyJKKHFYprArQmKHHbVXaXPBCgKvMQvnKfA47e2dsG0DQrUMXepCvvksgrISgt7OSYGhL7FiAvdsGsmNf5BlhqtnN+NO2cOL0z/ebS/knI0Lmzp3Lr3/9a/72t79x9NFH8/e//50XXniBDRs2UFJSwp133sm+fft45ZXIImB//etf6du3L0OGDAEi84bceOONXH/99bHg1GXLlnH88cfz8MMPc+aZZ/Lee+/xhz/8oVNDdHtChAgRTml9iJiK0+hxAaEqhANmwgE74YAZNWRCDZoi/0NGRMiAGjYgwnokOYzeWhPpoDPK0RndKfKWCbpzcDf2Qm9oIiyHEHKQsN9OyJtF0J1NyJOLp743UsiG6stEQkJSFaQUY9b3DwLZ4MZgq0ZvrcJgq0ExOVEMLnTGJhRjE4rRhWJ0ISsJbJ9dLVlAOKwjGDTh89kIBk1UVfbDZmtADVqQhERW3g5Mlga2bxtPQ80gTp/RnxrP3UB8kZhviwjuZCJl8YfLWbKyechjPAFxZMSknCyNAR1BVaCEzWTpIDPswBs04PZZQDWgqAZkVU97ASH0Ljz6OgLGWnKDeYT6rSa/YB2hoAGf30ZdbR9crpw2x0hSGJ0ugNnciMXiRFUVTCY3RpMLo9Ed+Wx0I8upr39VlfF4MvH7rASCJoIBM4GAudVnE8GgGVWV0ekC6HQB9Ho/Op0/8l0fQK/zozd4ycyoxmJtTFlmZ2lo6E1m5t7mc2/ZHv2NhZCQJNEtcSWERGNjPj6fldzc3eh0oTZlRDqUZAutRdDrcwgGa9su9wqx71E3ZLI4mVQvWFlZx1Jfv7TrJ9uMomQQDjeRaOis0VjI2Iy3afxgJ+HGlpEySqYBx+ml7JBX8/bbX2LPqGbo0C/Q6YJx8oGtW88i4K+n/4BvkCWVYNBEMGjE482gonwQHk/iibt0DTVkOLcw6KydGGyR30QNw8bXSrHk+MkdXk9G3+TP4ChqCIIeHT/MHYhQJQacspuMPpFjhYAdnxTTkHUqbm/8hecA7IQp6fUBWQOdsTy/XXQcLnM/AIpM33D25dd1Kg6ws/wkRsdAxGpRW1vLAw88QHl5OSNGjOCjjz6K+ZbLy8vZvXt3LL2qqtx5553s2LEDnU5HaWkpjz76KFdddVUszcSJE3n99df5wx/+wD333ENpaSlz58494HOE1NWtJCwiAkSoMlK7h23kxk1PgABIchidyYXO1H3zpRoyICnB2JuBJKsY7NUY7BFfYzBooLGhEH9YJqx4CFuDhE21yJnbCYf1hMM6wqFIZ2VQghh0YYxKCL0+gEHvR9f81qgYPIR9mXhrB8T+wj5HqtphsFdgyt6BOXcr5pydKHo3atgIQkJSAhGBoU9vwqNkBAJGfD47Pp8Nn8+GhMDY3DmaTG4Mho5vp5IEOl0Inc6F2ezC6czFanVSXV2Czxe52fZWDkBWdZjRYQkb2bBqGbacYRgzytFZ6mN5RqP3y+seRDF4UNotjKkY66ly3cnaZSpLV+xEEgoSEjZrPQa9H3/IiMvlQAj48uvFLZlKKkajC1noUANWQiEzkpARgEkWyEoYlxSmUVeBpAPJLCEhgZBoifiTEEiEUQlJQYQc6dwqjE3IdXaCweHY7TVYbXWU9FsX6/yjf+mIi3gIIeF2O3A15dDkysHVlIPb7UCI9MRuMGgmGDST5BmN3uDBkVmJw1GBI6sckym9DiIZDsfeuNtbfuvE73qhkB6/30LAb8EfsGDQe8nIrMLvt6DX+zEY/LE8HI6OS8G3XKOpBQgQESCQINhTYvOWBxGE2bLl4Zglw9kEdfVPtg0YjWOxin7vCQEC0Lfv5ezY8QyJpsPvZ7yBun9v7nBcuDFA7aubsJ5bw7Dhn7Fh/RS+XTedYcOXYDa3nSE1HFawWddT1jiI1as6MamlEOgaa7DVbaP0zN0xASJEJJY1f0Q9ZSsKcO62YxkpY+mrUlCwLTKyJwGyDrZ/1BehRs4v6G7pov0NBnzeYbhNSS5uwG/w4xgQOcegV2HXol74hJ1omE+4IUTjDjv5HeNbDwraAnZx6AlLyKa1r1FW/wcA1LAOETagGNoGc6khA0F3bvNfDkF3LmrIiGzwoBjckY7J4EbWuyPfjZH/si71lMdRggETblc2Ta4sXO5s3K4svF47Ol0w9hB2OMoxm7svbtLB57PicubR5MzD5czH587GZm3AkVWGPXsPJks9On38t5V0UVWZYNAYe5sJBY0EguaY2PB5I/9VNflQREkKYzR6ImZzkxurtZ6cnD0J28rtdtDYmI8ih2Jv4DpdAJ0SQKcLorR6Cwv5bYQ8WQS9DkLeLELezNj/oDcL1W9HVfWIkJ6wkAnpnQT1TYQMkf9C7lkLT2JUrNYG7PZa7PYa7PYaLNbGpJ1qWrmqMj6fFa83A7/PhseTicuVjdudhaIEsVrrsVrrsVgbsNnqMZudCCETCukJh/Rs3XokQsgUF28iFDaghvUIISErIRQ5iKyE2Ld3GJKkMqB0NYoSRKeEUYWEquoIhXSEw3qEqkMg0OmC6PU+dLogshxCUeK74HoaIaCqcgA7d46OBQ4CSAgEEpHg3hoKCzeTl78TWU7e7pJkiK0TEp/0rCUJcgcE+XmnUlX9YcrU3aM5fmXiEqqrP40TV1fEoIF/QP1nVhsLSHv29lqDe8T/o76uiA0bJgMSOTm76VX8I5mZVR3SN9QXUlY2mNra3pBsLhUh0DdUY6nZSelpu7AWRII9DYbISr+BQDVqUGLjawMJeXUIQB5/FC6Pj7z8HQwY8E1cq0zTXgvbPoy8hEuySkaJk15H1WDMiKRVXdfz1ZqGpC1X0m8tfftGAvsrvsmhYnU+7n5DUM2RWV1tP3xDRnZOt1bJTcVPxhJyqFNTU5P2FMOtsdlsBH2Zse+yEqLmh+moIRN1VWaCniyCnmzCASvQ/HZvcDe7EJrQmwQ6cwDF7EJSvOjttYiQF13QgKJEzLEBv5lgyEQ4FLFMeH0eZNmPTgkhSQKP147HnUUgEPVHtyaipBsbc9m1KxedbgyZmX6ys8vIclSQkVmJrlmtu1xhQl3s7yRJT2Zm2zdikyliZTBZtpOV2/aBGhLgSvFSGpHMBvR6O15vBm53Fh63g0DATDBowuUCv7/ljd5sbsThqMSeUQ00oNcZkMw6FMVCfX0hGRm1hEL6ZkEiYzHZCQeshIMWwv5Mwr5cvEKiMeAnGAqCkLBYGnDk7SQzbxtme+sgwFoslto2dQ0GIRSCTFPbm11ndKEzupAsuwh7BXpoMztDwOPA7cylye3A6bXjcTtQVR2EgFa/h9Uaf/bLUCgU59oV6JQgVqMPk9GL3uBGUULIcghZaf0XJDNToNeHMFmcKK1cWsGgoLExvpVDDUVicGQ5iKQkWCLcLqMoEqGQDrO5CSFkvJ5M9IZGHFnbKC5uRKf3EwiY2bnjcPrb9xIKeWlqfnn1+yV27hhJMFiEJEls395A/wFryc31o9dHfnOfz8L2beOBUlRVsHGDk/4D1mIwtLxB+v1mdu4YTf8B69DrvQgBgUDkL9ZaAgx6G9nZmfgDHWMbGhoinbmi2AiHOyfizWYJo1FGkqCgcDsFeS6+X308TX4zTR5Pm3k4ciQJfa/dOJ2pb8TSAedTW/ef6Bm02iPhdIYpKpzJvrL0YuNaYzBIWCwyCDoIkETPCAULYeGJHx4kQCfbQXFhtcpxXDqCwoKbqampQ5LGMqj0LWq2fEXAXYXBmk9u6TEEdrmpK98Ut746WSHTZKew7HDc/pvIytqMrs82vt1Tyr59RezalYfZ0kBh4TZyc3eiRK9XaR+9iveRl2ulvnwI/qABk2Mvel2APXtG4PVmk5/nRN5ehVofoN/0vTEBEghkMmjg3zAYCtj5w1I8oSoKRlfx4+JILElGdTlNqoWmpl5UVpgYMXJRG6tohl1h37KW6SWEKtO4w0FYDpF7ROT6M5lW4nYXJ/ydJCmM2bSJhoYwIgw711gxAao+MlNqZHiuoKm2hu3friWzdxpzqCQgJycn7kR7gUCAmppEwdFx6qxZQjoSVXFd5S9/+QunTz6f9TtPwmB0I0mw8qsLEcAjjz6F25PcnJaISZMmccIJJ8TdN3v2bKqrq+PuS8UxxwzlnHOO5/DRn3TYd/PNZXz3berFjuIxatQALr1sChZLI1ZrQ/PbbQOKEuaB+yv54ouumcFLSkq49NJL4+77+KMPWLnqmy7lO6Aoh9cfOwW3zsnmymLqa4ux2RrIszv5zxs7eW9J8mmoE5GRofCvV0cRChkIhwwougAmkwuDwcv77zXy7LO1qTNJwGt/vhU9EnpJRi+BTgJFgi/XbeK+f77f5XzfersEh6PjW9K6tT5uvbUszhHp8cI/ejNwQG82bz6cAQMj11tkmGWA31wR36WRDtdcczrDhmW1BAVLMjNnzqSuri7hWlTp8OijV3HZZWc3zzUCrTv2k6Zs73K+11+fw5lnZYIAY1Nf8jZdh6UxDy8Bjnj2bJo8XbNM/v73V3DLLeezefMDbYST0VjEb67Yw48/7k5ydGLOOCOD392Q23ajCvrdEjf+cR9r93btGXHE8FE8+Kg+7gi317/4gvmfpg60jcdRfUbz5i/+X4ft9ZKLPyx+hPdXdM1d1KePkbvOnoi7wkzfE8vIHhSJuwgF9Xzw7mj+8vzcLuWbmSnz/INj2Ls0MotIQekgBp10KvV1dSz64nOefPa5FDkk5okLz8A1eAwAituJZXfEfSWPGM/N997X5XyrqqrIy8vrsP3zzz9n8uTJAJol5GASMrnZtu0Ihg5bEnnLUmWEUJpNrD1HxHQcRFG67sLw+62EnX0hlIVQ6nvMDO1wVDJq1KLY954e9SALCaswYRMm8kQGvdUctoe+ZSVdEyG6gJ2iDb8FYCCgIgjWh6nzllFg/qHL9dQFbZQum8UeuYa9ch1NkheP5CckB9i1axnwWZfzzjv8nQ7bBBCu6Tn3mhTWYWocQMa+4+i91gjc2uW86jdPoWLXieg8DjY1Wuk36CvMGV0Tz61xu7Opru4HgFUYOSowCPM2I4OPGtytfAcMmEJ+/nRGjvhrB5dAdzE1DCB36zlY60bEtlkwohddn1LdZOqFaZ1E/iwDvuJ+qFkG5PoApn0qUtfefVoQoNsLxh9ljD/IGLZJyH4JXW3Xb2qzP4sBX9yH07YLpwiRrxcYGgbi8Qpq9nTf5RNxarXUL0vYcARykxyRIr+QhBqS6HV0VUyACAEVGwtoqup6OwgB5ata6lVpzGDrFxGhtGHrzi7nC6AaWhYqlAMt1lFTAivqgUYTIfsJt9tFbW1fNm2cxIDSVeh0AVS1e/43q7WOkpJ1kcBJswtJCrFn90jq6pJPYJSKIeFizvdNRP5iPH7rXsI6D6rOC5LA6Pob0PU3PknVofNlo/PkoPflobgLMAYcmOvegi6KhRzVzkW+YzBjRG4n6npS4slIGNFR5OuLtak/kN7cNfHIFBYyw30ZEW75rQKECHrreL8bIuRAIJQQ3uzNeLM3U9Vk71Zenp3H48npjQ7I2XUa3t2nEZDC+Bq3AVd3Od+jvSMZ6R6GXglRQCaSAP9XZYSPSLS6aXo0ffEdnH8++fnTyc2eTMXK/+Br3I0psy9wWZfzzd1yHiWWX7bZpvoaCWxdiAh0PVDWv20HlV9uwzTqRizm7MhGK6g5dYRWXtLlfK01Ixj4+R+QfUZEyAfFPkSBDxH2IX/4KHi3dC1fGTZ6BfuckckhfwQK9YJ+BgldqGvBza2RTAp7/dVsUyrIVzPpR07qg5KgGMIMPndnh+35pRXUvt71vNWQTNgf6Y6DGdn4VKnHHmaqoSXiPTpRmT0nl8zinpuQsztoImQ/YbNGHn61tX2pr+nDdAZjQOEpdKSeazA+mc5S8vZNQtEFUENGgr4M+gmZQeh4KWQFGrqUr4yEDgVCCubGgW32KYH/JDgqNbaqsQz69MG464jqgl1zbQAEpBABQuxQqjEIBYs+hCSHCAmJeqnrD/CgFKZR8pAhzG3enrqLQPCN40tyem1EMTciJBVJyAT9Zsp/bEqdQTJUkBtBqZXQ1YKuSkJ2S1i2dm+BsqznFOylCqFiQShPEMoXqBkQpj71wUk41qZjcEb7gGAdP3oSz9CZDtmKnr5K5J4LqgK9LGECKleUdytf//YAajCEa/FnVD7yKCJkRzJm4vJ3TUBHkcMt56t66ghsWUBw11JUESKsKBDuWvCo99t9GE64ARRTm5miJVMWkjkLPF0bnqz35qMEbaCApOjB2CJGJaXrb9SKJDHAKGOQBFUhFb0kIYDKIJh6Yp2oITn0XqcSVMOskHZT4XTEHezbVRp2HIVr3zhCvkxUdXGX81GFHk/fwUhBPyGbo0ffplpbQqRmS8jkS65k3e4ky+geQLSYkDhEY0K2bduG3d75Nz+bzYbY6+Wvrz6Pm8iPfrn/RGQk6jwNXb4JzDojFkP8h3W914maxixfAoGfIEHC6FEwosekGLEZLXHTN/qaCKldeyDqFR0Zxvhvok6/C6/UQMhcS8BajsdUhagvoajxcPaY61gSXg+oZGTWoNf7CAZNOBtzAZkpxtGMsg5km2sPX+u34G49MZcPRvv60FfkN0exxo+Kk+2RGTDVphY3lizJZJlb/JeyVY/qjuz3BLx4Q/GDlJty1lJ72NuEjC0PeJ03k5wfz8JWPxjV2YDPBhukLRhMZegMAXxY8HuLGRLoj1sNsb74K/oNWIfB1OJb93tM7NsykiG7D0caMRehusAPsldCVwX6XRIF5XrkYMdzDKgqLrXj9aCaBaoVVKtASKCvlJC9HY93KApyO9+ZahAEEfisQ5DshcjWXEIFfsI5CnqKMTaVRIZRJ/G5ZZps6OSO7z4hNUSjr6MLKaAKAgJCQqBIUizuRdeuDLvRikHpONopZJBobEgsnBIt6SDUEKq7BpveiH1AENfCzzCOnIkctS4A1bU7CGx6j9zrL8DznR21KYhAoOo8SKoBWU08y6pFb8YYcBLY/DHB3ctBhDH0H8JenZXPBvfFI7Vcl5aQjvFbysira0SEfOgHzUCXUxo33+gzQnVXE27cC5KMklOKbLClfEaIkB8R8kPYhwj6EKEAqs6Lagmhz7RgtzuQMGPw5COrLR1bo6+JoBqZWwivn1DF96i1W1BdFajuGpAV9ANOxNDvONCZ8AswSpHlNVI9I4Lh1MG45iMLCGyqI9zqXjY6LJTMPJw9niDSW1vwhQVfe8IEBBRLPvrrI880VQj8Ajh9AD7xNZX1z6JrdR83bM+nbJkDNSyQDSpZpYJ+4wSGjEpq1p9F/ZaTYmn9QS+BBM8IIVQCrv9AOPLS4RowHHQt16skSVgs8Z/BwWCQQCCA1VrPyFGfAmCxDGDc2Ndj1255+Xo2b7kcSRIEA2YKrP9k+WtfEfQsRenfh1BmxEpj2b4Dk24ip15/Nr2HZXZrsdZkgal79+6ltLT00J+s7FClJ4boetZV8c0bX7JI/z0Av/YfjxE9AhExwxPGgIIQKnWikbAIEFYDICno9DZMGDALA2YMXXorV5ujT2LHGhQIJBYTrTvc/Y8gjMq+nDX4DbUYAzkU145FQUbJNJF5an++eX0JX+u24JZbrV+iGjkqNIgjfnECnnXr8HxvQEVQqTTixY8ZIwXhTGQkFHMZYW9k1cjWnUz0creMDGAZPZraV+NH1wNk/2II9W9vRPV17FiFGo486Bt2Irz1qE3lhHw7ER4nUjAcGRaTztSvrVAlibBeRigSclDtkjlaNQiC/QWBAc3WC3tEdITtgsw3dCBD4/kh1OjcSwL02yDjHR3GcjP4/ah6EVfYJEMyWJFshcj2IhR7EbKjP3JGEZLeknItqNYIISIDgETEohFu/hyi1cRfCIICDFLk6pYkkAR4BWTrJDKUno27Uj21hKo3ofqa0GUWIzv6IhkziApc4a1GtqSedEEEfYQbdxOu3Uy4bgfhqg0gVCRLLsaRM3GcdwSKsYnyRx6lyl6Ez5qBye0kv6kc+5TzUN0jUd1VBLYvRnXXou81GjmzL7K9CKkTQy1F0EO4dguhms2Eazajuqsg5I/UxeRAyR9GsNRI7aSvCeS3FtbZ5P1wEYayUbglPXUhQX1I0KQKQgJsikSGLGGXBZbKdejXv4Yu2GLpa7L1YcuQX9Jg64NJgmKDTB+DTGYav1dAFXgFOMMqVkkiW9+q85vRj+Lje+Pf0YjaFEC2GzD2z0SSJfb9WM8X/28tjeGWgckZMhxn18WEbE1IJe+3o/A0bOKzOX8D4x50lhDeGhP+xhZXhq2XG8eAQdT+cDYi3G5SH1SQVBCJnQuB0HeoTREREbLY8fY9DCkURAoGEAYjQpdgygDRfIEDow5fEBtaPPrwOeTkRIKut2ydxe7d/wBg587Dqd0xFWv1YIRQqc9eSdgYAAE5FRORJR22LCO/fnjiflmVHLQhuocEslVHfzWfKcGRrJA3sW/1n1D8HnR+N4ZgCGMgjD8cGQ/Y/j0gqNNRn5tHMGsYetsgdIUjkJRIh9souQkQQo8Oh7B2iIkQoQCoIWRDW1VtP64XTYt2NweHtu2UJQkcZ5Ym7HBbp2t/fJv9xkiapHnIIYSqQ0amb+34NvsEYOjjJFSxlsKv3+X8kedTZZVjAiPfpRJc/yaB423U/eMxkIowjrqAXuZshBpC+BoJ16/Dt2U+auMuJEsusq0w4mMPB1ByBiI7+hEqX4dvRQXZiz7FPMKP+xsfsrFlNJTqb8A6zox5eDblf5yLvv+ZhOq2otbtQG0qi7zhuapAdBR13bmlZSGQkwjFeISzJUJGK8ZyF5LNhr/USePMViIDkOsg8y0d1t3ZqI2NmL6TCZSqhDMFSqOEYZuMJCRyr/stNc/+BTkooSJw9ldAAUO9wFyrJhXDIuBG1G1DrdtGm3dXWY9sK0C2FyHbC5FshajmTPS5JciqocN1IkkSOiFQEdibO6eOIlLi22CIIXodxlb7hRDsDQq+9YTI0cn00ss4EqzFEcsv5Cdct41w7RbCdduRrHkoGb3RFY9FNkYenrIlB0PJsQihEq7+gXDVRpS8ocjmSCNLcQSICPkjwrTmR9S67YQbdiG87daFUQwYDjsFfelJEHAhUUfGtFOxT5lC79XfEKquRpeXh+WIcbhXrabutQokSx6mkRcggl7CdVtRnXsJ120FSUFSDEimTCSTA9mSE3GbQCRtwy7C9TtRG3YSbiyLLN/QXA1dzmEo+UNR8oYiWQsICoFdlrGtnYkvezNuxY3LmY+roYh1YagPQ7y5RnwhQU3Uzpt5OEwchTHoIqNxG5Iaoip/LEgR8eATUF++nty9n2DM7oe++AhkWwHoTLHf06cKdvvD7A4I3O1elUuNgiFGiQASWHRIsoSp1NGhTp6mAHWtqpqjSEywKegkKWYFy9XJeL/bwPtvPdqcqqN7KaPEQ8lJe1B0u3EMWI2nZiCyzofO5EQxOhGqjrIvnsDnSnz/6pWRBHRrEKE6dJ4mdE0NGKr3oTTHafgKh6LXDUUJmwkrfsKKFyQVs6sPTVk/ENZ52Ld3aEyE7NnzT3JyjiMc9lBWFlm1WlUVKsoPI6hUotcXYghmouojLzOyakSWIl2+q95P+ZYGigcnngn2QKFZQuLQE5YQ19crIg8NUxYqAvf71yB18s04hqQgO/qi5AxCyY6YYYWvHuFrRPXWI7wNqL5GhK8BQpEQeMmSG0mfMxDZmo95tB/nu8s6mJRVTx3+9W+Q//tfU/P8++gHnodw1yC8tZGOJeRFBLyRt6e6zcjmXETQiwh5IBwCRYck60DWoXqqkCQJOXMAyDokRY+kM4HBgqS3IuktBHZ8hoSMYdhZKPZeoDNC0EO4cR+BH+chQlWgqqjNZkLJmAGygggHIBqw195nLilxBUEqrMcfj/uLLwAJJXcQkjET4W8kXLMFEJhGj8a3bl3nMpX1kad7uKNVSQBBey7ojEgBLzp3XawjCCgyfp0OnaSgCAjJEgHCmIMhjKEwHoOO8kwrYUXGmmtGytJTb7awu9KPEDB2VyUjLr6cmr/8BVUS1A7XEcyS0NcLcjaEkIVE8TNPA1D5yCxCFS2jPHSFhRTcdSf2KVPYOuUk9nqb2NgrF1+rlXVNgRDDymookvRUhfzszM3EGgiSEzZi9QUweJwYgoHYuTSajTRajDSajaiyRJ7TQ68GF4Zwyz0gWfOQm60nkt4SaSGhIoTAXbMVp70feXoi9kMhkGSFoKRQGZKxNqxHb87Gnn0YyErkGhQqIuglGPSx09yb7RmHYdLpKdLL9NJLZOtkQuEQLk8jTq+TRp+bJr8P1WgHcw7CmIEsy2QrUBBqxO6tQjLaUDKKm5dYSIwIuAnV/BgRKdU/oLoSjKKRdciZfVFyBmIonYJkcgDgW/k3ej16I9YJR8bPPxxm+3nXYzjswkjbxbHuBfd8hqHvlKT1BKiv/BZH/qi4+QghWO4JExRgkSUawwJPiseWXopM6+VPsycxe6oYtO1tcmrXd5C1sqMEw4ATUXqNQw35WBO2cIRZpjok2OIX1IZbCrHJYJLh2OvHUDTIQfmWBtxOP9YMI0WDHGz6qozP//Njm+lSshUwyxI+ASYJjrA2B4OKAPP3vognFN89kTkgSL+Ttja3Wcv2aO9Z+931jDv+Fyx4oePq360JB7YRdL8XOVaSkKIZ6PMxWn+Z0GoY1DXRkL0OWeiYdNLHhMKRWKejJiygvn4FP26+N5KN/gQ+WxQJONUFbWTWjaC24OvIPr8DR/2oWJ5TrxjGYeMLk9a3q/xk1o45VOkJEdL4wYdU/enfmI6MRPy7ProJgp6I0teZkfQmVJ8Tgu6ICTT3sEgn6KsnXLsF4etaAFlComYMiJiTDZkIXy2qs3nOB52OLs9K9j+GaJ45UmqzDeosRvwGPcZgiGy3L7ZfycpinxpI2LEjSazp2/w2HecJN67Bx4YMIz5Zij/GWQhMquDq1/7L+pf+wdKFH+BTWjpNU1jl2GmncfiV1wAQDgbY9t5/cZWXYSvqRemZZ6HoI/EL3/59Np9++mHCupw4+WSWLfoobl0kVSAhUOP4iQEkIchzeiiubyLf6UHZz48evyGT7f1Pp7xwAkgy6a/UFMEogVmGwUaFXCmAIklIuogZXoSDhOt3EK5cT6hqI2rjHuKtbyJZclGy+iPbCgnXbMY49tco1hbLSfQlALWMgYs+RVIUVFV06FBlWcK5cCGVj76c+EXi9kto/NKEFExsrRR6WOQOkhmUGGnRYZCgMSyoDQmqQpH/6fwqsk7CLgTDTQq5uojl069Ggku3+lRcsoQaitMe4SCOxi04GreT37Qda+MOCHec8VQy2kFvZfUR12HRZzHSrMMkwXa/ykaf2uZ3LB2XT/nWhuaVtyPoTQpBX8uLSUGuSn15EwF9i8XTGGzkxNIsDM6IK6vat5fF5f8hUQtkDx1B0RGfJlzJvHRMPtvWVvHl3C24G1rcyLYsI0OPKWLTsnKaan0Emv6DCLedft9g/yWyLjJRmSoF8ZnL8VoqyKwfgS4csWq7rTuxmRyccmUd27ZF1pvKzTkRZ9N6AoGIdeSIce/xyktLqG2IDHs3eQrxWSKC2ODLJqNheMyiedZNY/abJUQTId2kJ0SIe8VKdl9yCbqiMRhHXYCks4DOgCTJkYfG93MJla9Fycoi3NAAgtjbuOprQK1tHvKm10fiC1IiIVmymx9OgnD9zsjKRfuRRGGfhxJRcRDQ67D5Atj8wU7XOShLVGTaqM6w4DQb8Rh0mIJhhpXVUNjopiLTmlBg9Dbb8Vx4Hp9+1ryqc5yO3aDTEwgFEwoMRacjnMZoiYHjj2Lrqq8T7j/j5rsA+GzO33HVtcxoaMvO5cRLr6R0/AReuPaKNvvaI8kyIk7Aa2fRhVUKGl30qWsiq5Vg2x802XqzpfRcGrIO63IeEpCrk+hrkMhWZMxypKMXQkUE3Pg3/Rdd0djIy0PQjWzNR87uj2SIBLb7Vv6NUPna5kDRgS1Wt9qtgKD4mafJmDYtbidmdRg57oJBlI7JZ/1Ln/D1Uhc2swOTFHFruLwNHHWsjWGXnMTHd33FSDWMW4W6MHjUSPxMUET+Nwjwh7v/yD/ytP6s/GAHADk6KVaX2mbhMf70fqyatzNlPhP3zMORNxDhqiK4Z3mzmGtByDKVuePYUzwJQ84ATBK4VEFjKP1Jh4YNUin4x+9AQINjIH5DBsaAE0fjNiSdEcsZj6M0x3h8X/8FGxuWA2DXZ+MK1sdEyRFnXMmO77MQ+vUopgbCPgdyaATHzhxM6ZhWwjKBiFw57yuWv7MaSbYTdLXM76MYhqO3TkcN1xP2ryFk740nux5VCaAL2nDUjkFqnsh/4jkDGXViNl8unYCqtg2ClSQ9I4Y/g7NhIK/+518d2kEK65CQsTkHkmMpPmRiQjQREoceWUU3HGbrlJMIVVa2ERgxc78EuoICCu64nX033dx8UKufovkGy73uWmqe/UtkNxC05yEMFlBDGOr3xR7eFRkWNhbnY8/oh1mx4Qs6kao2c1hFLfawSNvKEVRk6qwmmkwGAjqFoCIjqdC7oYlst49qu5lNhTm4zQZEs3XF4g8ytLyWAmfL4GMBNJoMhBQZYyjcofMXQFCWYypGH1bb7G8wG9mRl4nLZCAky4RlCUNI5bCKWgqdnph4E0JQZzXh1+vaWiBkmQq7Ob442FdNkSojXK5YXaRW9aLV96qCHFYXZLb5TVr/VgOcPrZnmBLunzjxBNZu+BavsyGt9t+fGK02/O7Ek5gNOnIiW1YmXq58fyGrKoZQGFkIJCGQVYEurKI2m6ulaABqq/2l1Q1keZIvqdBoL6Fq9FWMzrAjZB0e5NgoG0kIZAm+84apar41shUwSNCkgjuFzjJJEVGSo5OwyhIVOxfhLTqBkRYd5lYPdo8qWO8Jkb/zTUZddyaVjz4W1xUWFSDzn09szh89tQ/rPtmTcH/JyBx2fV/baYtPZ4kGNe74tjruW/+xMwfR//A8XrlrGe4GH4lGqBmDTiYu+wOSECi5g8CQgXBXdhAiURoz+lNWdAw1OSOQ1CCZzp1U541JKkYMZoVJa+8nXJFgqLYk4Rl8JHmDL0OWZFShsqjsVfrYhjAk80g8ISdrahexz7OZmfc+QvHQkXEFRjosfuU91nz4ApLSD/AjwuWAAZ1lGmpwI2owMh9T6ZG/or66Nw2eKlQ5gD6QgdI8Gimr0MLk3zay8cdrEpQiEdj+HOu2bMZvbjsJoBK0EtZFXNqTjpzO5FOPTqveXUETId2kJ0QIgHPhQvbdcGOLqT7aUXr8SBB7+3EuXEjFI7OoamqIpcm3OyhM008vWax8k9cc3hqnI5w0fCzW196ixmaizmrG5g+CEHiMegKKQoHTTY7bR1VRHqvz7Anz6V/dyI68xB3yEQ0+evnD7PW5Eta1sMmbWByU1VBsslFuVFjtSNyxH9HoZ8xNt/LNH/+QMB/blBP5YuuGlG1SkWllc0E2lkAQIUG91YQuLBhWVkNBo5sl44fj8XsTWikOyCpnP0UStY0QyIqC2h1rSrP7aWq1h3BVVXwhKknI+QV8OfB3ZJszE4qD8qAaC5aMh84oYwiqiObRN0kq1axmJRwKWCTwI0UsA0JgscAlfzoREQ6z48OVuCqd2Aoy6H/qkSh6Haoqmjvt7q8Ovb+ZcdWI2Jt/ord+IKWomjQhjPL4DZEv7V/AhMA8bizeb9Z0OC6a0mfKYd3Ia/Fa8hPeh3I4wLg1T2J3J54TQwCVJ17FoIxxAHhCTZgVW7OlK1LaGu8iTn/6XiTkuKNw0mHN/GUsfumR5m8GJMWBEEFQ2w4hn3zZXYyednSsXc1WA8ve3UrNnsgLRN7Qr8gZOSduGY27JlC+4jeEZR91easjI3aiJXpzCZgjVs6MjAxuvPHGuENsewJtdMwhQsa0aey46bqEPvqMadMAqMy0sXhoX1x1LVHZtuwcTsy0kaEoeH5xPms+7TiFsU+vsKakAKPJDP44azc035hrKvcwuKSY1ZnGNtsBEIKdeZmMa/SzqSgHPK6ON3TzQ2FHviN+59K8f2NxLpkTjmdN1PUQp67DCvuwsaLjm050v33SVDas+go8noTlbCrOpdBhY02/wrYPr2g+/QoxVe6J/2Bq1yZrmtvEZW6Z0yEkC9aUFFDgD+MJ+BILjQMoQEw2Gz5XYiuG3mQm6OvuvNzpYc7IwJtkfgGTzY7P1dTxWmn+PnrG6az56L2uV0CS8CkSdeechuc/r7MpgRAdd/edHLVP4rMvNrOzZgm5RgdmxYY37KLG34DOPImRI0rZmGSyzzHT+rJq3k4KdTDaqNAQFtSEBHUh0W58SMtwk4ZwdNpAEauvxwvffLSTDUvLmoWGAfBh/XoFx10wCDUsDpgAMdn0+FyJXby2LCPHnDeIpW/Gt3K0dj3IspQwrqB0TD4zrhqR0FpSOiYfZ/HTHQOlCwoouOtORCAYV4REryizr5ajVz2A25TL5sMuoj57SGSHUMmvWkNx2VIynDtQRHIrcLnDxre1n5FtKCLH1AuLrtVEbM1CZEzOSfjW19H44fY2q/YqmQYcp5diHpF6KvhRJx3JklfsqOEmIIAId1zFV1bsjDrpyA7tepJjGG8+sppwSKV609GYC5Zhyd/c5li/s4CKb34VqZdqQgmZCeubg/kFKKGWEZNOp5Ndu3bRv3//lPXe32iWkDj0lCVky4plvP/nRxLuj/rok6WZduX1LJ37Kp7G7s1UKUlSxJe/n9/oW789JEjQQTh0BVlRULs4q2QbfgLWDHtOLpMu/g0fPPVowjQTz/8ly978d7fLMpjNBLyJxUw6dWmJPXkeV13L4nz2nFwmX3IlJpuNNx64q9t1jZFA7KRzf51x813IhoFpuBX8CKGSLdVikkJ4hY56kYMkyUiKhOiBOIueoH1AZnuiAiPZKI6opSOZlaMzpMpHhMN42g1JlhQlFleXLvWZA6nJHUVB1WoymtJbqM+rV1g6pB9BGWw6Byf3/i1yipFQ8cj51dC0hMiXr3/Mynf/mnD/kWdfy3EXnhx339pPdrPs7cgIHb21hn7T7kXRR8SkGtKxa9Hd+Bt7RxI7tlJtKovNMaIErZg9xbgyW4TLueeey8iRI9M9xU6huWO6SU+IEFUNpwzy05siUyuHOiy5rvFzIm3LQQLOuPkuBk2YyJYVyzoElUY79nSCStNh4vm/YNmbiafqT6cugyZMBCL3wL5NG3A11GNzZFE8dDiyrKR1b/QEOoMBJCnp/WXPyeU3f3kRVRV89+lKGquqyczPY9RJR6LTRawr29ZW8eGz7xL0LAbRyhol2dBbJnPUuVPTCsI8ELQOGI1HVGAkGsXR3tJxMGkbV9exmxJAveMwMhu3J7R2eI1ZmFU3BDuOwBHAygFF1NojFoK+1qEcnX9Gl+qqZBopvH18Wq6ZL1//mNXvv9JsEYkgK3aOOOPihAIEImLu7SeWULUj4mLJ7P8FReMjAajlq35N447jATBklOEqWE5jUx6oChZPL4zefFQlQGP297H8Lrnkkv1mCdHcMYcA+zZtSPmQDfq6tvy1xoFDbzIl/Z1SjRZJx3Iw7crrgY6jVtp36oMmTKR0/IS4HTvAiZdemfStP5XYsefkMuGcC8jt06/bdQGQZYU+w0d1KEeWlW7XNdXvAhAKdOx42tNUW8OKd+by3aKFbc531XuREUODJkxEDWwl6I6zpLxwEXTPIzNnBFaHsdvuFJ1BJhRIfC2lMiLasoyMO6Uf2cXWlAKjdEw+/Q/P6xFLx/5CUhQK7rqTfTfc2PHkm4OLQzozXx7zOFkNP5JbuwGbex8eSwH1mYOodwxi0La36XPR8dT8JRLc3zqPnbmZMQFiVuyMy5nW5bqGG/34dzTGnTCtPcddeDJHnzc1oehNhCxLTLv8CP5z/5eoISONO47HXryOcNASEyCS4qf/pLf5dv3RIDwghbG4+iKhEIo6sgTYLTZKSkq6fL49iWYJiUNPWEI2fbWEj/7fEz1cs66R6k083XSpOtx0y+luPXqinHTzSGUZOOL0c1g9752E+3vCctAZkpUDqV0TPVmX/VnXsaec2b24kjQ5/cY7WPzKP5K+UNhzcjnxikdZ8MLGbpU1akpvvlu0N+H+VKNj0g0Y/anhXLgw7gR7+XfcTtWjj1EWLmRL6Xn4TS0xFEZfHYO2vU0vpYKBiz6ladGiSPC/s54sjx+3ycBXg4pjc9qcWPor8tTibtUz+8LBWEbvfyvSglfeY+uySNyKYmpEDRkRoUggf+H4fzJo+Ok0fWrik+Ba9EJhSOOR1AX0+AlTk/cVSDBVP5aJd52edlBtZ9EsIYcANkfPTQLTXXP+SZdfk9aDNNUb+7hTz0ra4aZTTk9YDtIpp6faJB3LQK9BQ/ar5aAzpCrnjJvvSlnXnqrL/qyryWY7ICJk3jOPpYxhaqqtwWiqShiEOfTYorTcNQNG5dFroCOpFaNwQGZabpRkAaMHmu4K2oxp07BPmRI3bkSSZUI33EhuzXfszh+Ax2jG4vfSt2o7MoKCZ55GUpQ2wf+jdlWyI98REyDjTj2TIWNPSrqOVDoEyt1YRoNQRZdH0KRCVQVlG3JQjNWE/ZmEfS2Tr9l7r8JVNpodW4oZh2CKNBIHVrLMFqp0KsvdYBUWjgqVUuLPSttys7/RRMh+onjocGzZuUk7OVt2ZGXD1sF77ekpc74ky0nfLKPp5BSdVKoON1U5qYRMuvVIVU5PtYksKyk7y3QEBhyYjj1VOenW9UDR1bqqarhH7q+UpGkodjXUM/SYUXFdHAAbvyxP6q6xZbVYK5K5SQ41N0oqgRHP2hWdHK+16E2FpChxp7RvOwJRAJG5irbnlcRGILYfILCtIAu3KTIazhgIUVg6GPOIXHJ+NZSGedvajX4xknlq/w6jYuLhWrKXwO4mQrUeVGfL6KN4I2i6KlTKt0Rnhs2k9QxHkuKnae84QCZTHwSrjv6iRZTm6SRkYIZvAlnNIzXVptTuygOB5o6Jw6E0OqYnzfnp5JFOPt0tp6fqcSDbROPQoyfur55i5r2PJBWYqebMaO1KOVTorsBI5/fp7j2Wqoy47rTm0VOyKjhiexmbjhzFb/76T2RZSSgOvOtrum0piY6g8a6viSN20hvqu3lVBZ+8mNztl6OTONbW0b7wmTPIMLNCYfMKxLm/HbnfLCHa6Jhu0lMiBNLr5HqqU06HA+HrT6ecnqrHT6lNNHqe7txfI0+cntaw5lTuUJ3RyP/9/VUMJnPSfA61ESnJrvvuCox0Y2l+85cXu3yvpTPKStbpUBPMFj2krJYB1Q0sL+3F1MefSmmljC8ejDhOH4AaCFP/xuYkR7dYVer+80PCNKmG+u77sZ7/PrU2aTkAZ2brO0ybu9odIkcn0d+odGo0T1fQREg36UkRAul1clpHqKHRNbp6fwEpO7F03KEARQMHc8Ytd2FxOFKI70MjYDSZyIDk1qPTb7yDz15+AXd9YjdXuvP4RKZCH96lZ9+eDd91eb4ZQzDElI27kIAduZn0f+wxhh4zKeVxiSwlvm0N1LzwfcrjUSRIMqdMVBwAcctJZ2bdwxwGhsZZhO9HX5iwgGFmJe15TbqKJkK6SU+LEA0NjUOTdF0G8Tptk81O0Ocl3PymbbTakBUFr7NlBez28Q+HwstGqnNOFdTdU4sYQmSU0+avv+pSzEh3RiAq4TBTN+xEFpEJy3L/8yp9R47uUl4AnnVV1L3+Y5ePb439pL54VlUkdNckc+0ZJZiea0QKRn4fyaQgmievKwuoVEgSUy8btl8FCGgipNtoIkRD43+H7rhDa3bv4r9PPEhTTXW8rGOkWsG4p+OPDvZEcT1BqpiRdC0hJrsdX1NHUXXE9nLymyKBrH1f+w/WMWO6XNe0LSHdJGrBSOTaO6HYgrI3MqGeZWw+jnMGUfbHZRAWNIUFPxTaOPuWsfu9ntoQXQ0NDY006c7opvx+A/jFQ0/ywnVXJIw9AFj492fjWhdcdTW8/+dHeiRIM0oyV4vRaj1gAiRR558ui1/+O6XjJyS0FGUWFHZryH9FpjUmQtyLFnVLhBj7Z6JkGpKOoJEsOoQnvdXME9EwbzumYTlxR0llhcLU/SsSPCtbdWSeOgBZJ6PPtxAsd2OVwV13YNaX6gz7Zwk9DQ0NjZ8QUYEx9JhJ9Bk+qlMukvqyfUkFCJDUvQGRDldVu78WUtTV0l5oRMXO+3+e1e0yIBKsmwx7Ti4n/ebabpXRVFvDvk0b4u5z1dfx1kN/SOkWmnzJlQw+6ljOuPkubNltXRDufn2gea4Q58JPkq95lQJJlnCcXpo0TdZZA1EyDUnTpCI6Myu0zAVz2PhCikrsNL63LZbOcVopilUPgK4gMiusLEnQFECoh5bzQ7OEaGhoaHQDV0P3FpeElg63z/BRXR5ZpqphPpvz96Tl+N2JV2JOl3SCdVPN9zNowjFpTTYXbdvW56zT6Vj6+r+oLy8DwJLpiKxU3Op3SHeywD2X/wbP118T3L0b/48/YhoypCtNApB0rhHH6QMwj8hFkqVuD/WNN79H44KdhJ2R7cbDsjCPzovt0+dbiNo/rAI8TQGs0RXVDwE0EaKhoaHRDXpqduSmutq0hsYmdrVY0nK1pBq5kiowNd0JBSFx579v04a0RMimLxcT9HpZ/vbrcc8tI6+AC+6bhS07p0vuNPu0qXi+/hqApoULuyVCICJETMNyEk5ElkyoWMcX4Pw09eq/ocZIHEh0pI5/RyPu5eUASHqZrLMGIrVaVVrfbAkBsCsSrnr/ISVCtMDUOGiBqRoaGunSU8GeRosVv8edcH+qdYr0RhNBf+pFMVOtu5MoiLYn5y7qqTY76YprOHzaKV0+PlhVxdZJJ4AQGAcNZMC8OAsV7gfiDfUFqHhsZcqZWQEM/TMI1XhRm4JttlvG5pM9c3CbbcFqD5V/+gaAfQGV3F8P3e/z0mijY7qJJkI0NDQ6Q3eHvaZFqmV002TmvY/gc7kOyIKKyUjVZjqDkVAg+crE3Z3wDGDnRb/AuzYyAdiAjz7COGD/LG+fDj05M2sUERbsvWcpkgrOsCA4oz+HT+nT3aomRRsdo6GhoXEAGTRhYtIF9yD5BGCOwiIaKsqTF5KGAEk2Q2i0PlExcSAWVExGqjaTZJn3nnwoaR6tY2m6in3atJgIaVq4EOPVV3U5r+6SdA2b0/qjekM0vLuVOHORxYiOoIm6gCRFQso0Qb0Pmwx7ag+tETKaCNHQ0NDoAbqzKnAoFOzyxFutGT3t1KSuluiijHDgFlRMRrI22/TVkrTy6G5gsH3qVKoeewyApk8+IfcgihBIHlfi29aQVIBAywia1uvC6PLNhOp9yJJEsFoTIRoaGho/S7q6KvCeDd/1SPkDj5hA7yHDf1KLMiZqs3QDfrsbGGzoXYxp2DB8Gzfi27CBwN59GHoXdyvP7iLJUtzF5dJd+bZ9OlNvG64fI2JN1KeOGzqQHPR5QmbPnk3//v0xmUyMGzeOL7/8MmHad955h6lTp5KXl0dGRgZHH300CxYsaJNmzpw5SJLU4c/nO7QaXkND43+PRPORFA8d3mEei/ZIcvLHddTVMmjCRH771xeZee8jnPK725h57yP85i8vHpICJBnptEn0nLuLfdq02OemTz7pdn77C9me3jwj7dMZi2yxz4or2D75QeWgipC5c+dy4403cvfdd7N27VqOO+44Tj75ZHbvjj9M6YsvvmDq1Kl89NFHfPPNN0yePJnTTz+dtWvbriqYkZFBeXl5mz+TyXQgTklDQ0Oj08iyEls8LhHjTj0r6f54rpauTL52qJBOm7Q+5+7QWoQ0vP02jR98iHvFSkQai/AdSKIzsyZDyTTGRttE0eW3DNM1BsKo4Z5Z+6cnOKijYyZMmMDYsWN57rnnYtuGDh3KWWedxaxZ6c3sN3z4cC644ALuvfdeIGIJufHGG2loaOhyvbTRMRoaGgeDVOvYpLvOzc+JA3XOWyZPJlRe0WabrrCQgrvuJKOVSDnYpBpBE2+FXBEW7Ll7KTKRETJ97jwSe/b+ezH/SYyOCQQCfPPNN9xxxx1ttk+bNo1ly5allYeqqjQ1NZGdnd1mu8vloqSkhHA4zOjRo3nwwQcZk2RdAL/fj9/fMhTM6XR24kw0NDQ0eoZUwa3prnPzc+JAnLNz4cIOAgQgVFnJvhtuhGeePmSESDozs7ZHUiRCJh0GXwibDE013v0qQjrDQRMhNTU1hMNhCgoK2mwvKCigoqLjxRCPP/3pT7jdbmbOnBnbNmTIEObMmcPIkSNxOp0888wzHHPMMXz77bcMGjQobj6zZs3i/vvv7/rJaGhoaPQQqUatHAqjWg40+/OcRThM5SMJLO9CgCRR+cgs7FOmICmHhthLNTNrPNQMA/hCyJKEZ48LDuuZmX67y0EPTG09vSyAEKLDtni89tpr3HfffcydO5f8/JbZ34466ih+9atfcfjhh3PcccfxxhtvcNhhh/Hss88mzOvOO++ksbEx9rdnz56un5CGhoaGxk8Gz+pvCCV78RWCUEUFntXfHLhKpUF0BI1ldD6mUkdSAQKg5Jpjn/3l3V9DqKc4aJaQ3NxcFEXpYPWoqqrqYB1pz9y5c7niiit48803Oemkk5KmlWWZ8ePHs2XLloRpjEYjRuOhM5e+hoaGhsaBIVRdnXY6EQ5HREt1Nbq8PCxHjDtkrCOpMBZZUTfWAhA+hCYsO2gixGAwMG7cOD755BPOPvvs2PZPPvmEM888M+Fxr732GpdffjmvvfYap556aspyhBCsW7eOkSNH9ki9NTQ0NDR+Pujy8lInApwff0zlE08QrqxsOfYQDFxNhKVfBlH7h5zG+jQHioPqjrn55pv5xz/+wT//+U82bdrETTfdxO7du7n66quBiJvk4osvjqV/7bXXuPjii/nTn/7EUUcdRUVFBRUVFTQ2NsbS3H///SxYsIDt27ezbt06rrjiCtatWxfLU0NDQ0NDI4rliHHoCgsja/MkwbVoURsBAi2Bq86FC/dnFXsEe4kdtXkwrN6XeGr/A81BFSEXXHABTz/9NA888ACjR4/miy++4KOPPqKkpASA8vLyNnOGPP/884RCIa699lqKiopifzfccEMsTUNDA1deeSVDhw5l2rRp7Nu3jy+++IIjjzzygJ+fhoaGhsahjaQoFNx1Z/OX1PGIbWju1CsfmXXIzSnSHsWgw9N8fuawQIQPjbVrtVV046DNE6KhoaHxv4Vz4UIqH5nVJkhVV1iI4/zzqHn2LymP7/vyy1gnHNovu9/eu4ycQEQs5f5uDKZethRHdI2fxDwhGhoaGhoahwoZ06ZhnzKlQ+Cp8+P5aR2fboDrwSRs1UOzCHHtaNxvIqQzaCJEQ0NDQ0ODiGumvTUj3cDVdNP1FF0ZqSNlGaF5ATvfvkNjmK4mQjQ0NDQ0NBIQDVwNVVbGYkDaoyssxHLEuANWp0Suo1QjdfQFVtgeGcgRrPLs93qmw0GfrExDQ0NDQ+NQJZ3A1YK77jxg84U4Fy5k3w03dphgLZ2ROuZiW2yEDPX+hOkOJJoI0dDQ0NDQSELGtGkUP/M0ujgTaUoWC7bjjjsg9YhOMR93PIkQCCGSjtSx5ZpxNy+gq3iCCPXgj0vRRIiGhoaGhkYKMqZNY+CiT+n78sv0evJJLBMjK/gKj4fG9947IHWITjGfaCCxBEmnmLdlGWlqHporCwjV+fZPRTuBJkI0NDQ0NDTSIBq4mnnaqeTfcnNse93LryBUdb+XH6iqTJ0oSTqL3YCrlfEjVHnw40I0EaKhoaGhodFJzMOHY2meBDOwYweuL77Y72Xu0NV3K50kSwTNLbErwSp3j9SrO2giRENDQ0NDowtkX3pJ7HPdnJf3e3kVA7PxtRrTujcHWttfVKDGHkmXCJFpin32l2kiRENDQ0ND4yeJ7YQT0Jf0BcDz9df4fvhhv5ZX9PU2TM3LvngMMGumwidjWyJEJGDOVJk8e+KV6HW5ptgImUCFJkI0NDQ0NDR+kkiyTPYlB8YaEtizB8vT/4p9f2GGTLVD4rVJMg3W5voAdksWY/PHJszHmtMyQkat8x30ETKaCNHQ0NDQ0OgijrPOQs7MBKDxww8JVlX1eBkiFKLs1ttQ3RHLxZIREl8Nj3TfHiO8PKWlK7/iE4HkDyTMy95qhIwUFoTrD+4IGU2EaGhoaGhodBHZYiFr5szIl2CQ+tde6/EyambPxvvtt5EvxYX8c1qrrluS+GqYxHf9Im4ZXWUdNbOfS5iXLctEUyvrR/Agj5DRRIiGhoaGhkY3yPrVL0EXiRhteO11VF/3rAsiHMa9YiWNH3xI7cuvUPO35yM7FIVlv52A1xgRHL8c8kuuPfxakCRenCYTah74UvvSS/i3bImbty27xRICB3/6dm3tGA0NDQ0NjW6gLygg4+STcc6bR7ihgeqnn8E0YkTCheWSLT4Xb12YKDnXXcu/5TcB0Mk6rjr8KrJMWayoWMFqVvPuURLnfyUgFKLi/gfo+69XkNpNNW/LMrURIQd7rhBNhGhoaGhoaHST7EsuwTlvHgB1c+bEtrdfWC7Z4nMA+264MeFCeeuK/FQ3VgNwfPHxZJmyALh8xOWsrlzNfyfKTPlBR3atH8/q1TS+8y763r3biB2jRYdPryCEQJIkzRKioaGhoaHxUydYti/u9ujCcjzzNBBfZIQqK9n3uxuQMjISChCAd9f+CwZEPp8x8IzY9mOLj2VQ1iC21G/huZOC3D03sr38D39ok19U7FizMnC7A9gUCFV5EKpAkhNNBr9/0WJCNDQ0NDQ0ukF0Ybn4OwUIQdldd1N2513xRUbzNuF0JizDY4QVfSIr3zqMDo4vPj62T5IkLht+GQDfDpD5cXhmm3yjRAWRSXXFglNFUD2oI2Q0EaKhoaGhodENogvLJUO4XAh31ycHWz5EIqiPWCtO7n8yekXfZv+M/jMoshYB8OfJLtR4ho1mUSJv/vaQCU7VRIiGhoaGhkY3CFVX7/cyloxs6a7PLD2zw369rOfiYRcDUG+XWDo0gXtFCPQNZW2DUzURoqGhoaGh8dNEl5e3X/OvcMAPfSKiYkDmAIblDIub7pxB52CXzAD87RSJcAIdYvLVHzJzhWgiRENDQ0NDoxtYjhiHrrAQpAS9viShFBSgKyhImkZ2OCL726X5opUV5IzSMzoMu43VQ2/h/OyTAAjpZSod8Ysy+utpCoNods/4dzrxbWs4KFO4ayJEQ0NDQ0OjG0iKEhti20FkNH8vvPsuCu6+K2maogfup/iZpyNipRkV+PLwyEBWCYnTBpyWtC6/nnwThuZF7rb2iiNWJAmbXUGF2Boy4TofNS98T8VjK/Gur0maf0+jiRANDQ0NDY1ukjFtWgcBAaArKKD4mafJmDYt7TQDF31K35dfpteTT9Lw3N1U2iNq4aiioyiwJl4hFyDXmsepGRMBqM5su08QsX4U/fJqgDYuGYBwY4DaVzcdUCGizROioaGhoaHRA2RMm4Z9ypSEs6Gmm0ZSFKwTjgRg4bI/xra3nhskGcNGTubdr7+izi4RkR4Rau3wykk67tptQC+FaAoLivQdj2+Ytx3TsJwDMneIJkI0NDQ0NDR6iNYCojtpALwhLwt2LgDAorMwpe+UlMeE1TD/+P4fIEnU2VsEyOKREn87RWak9zD0uyXMEm1GyLTJo9GPf0cjplJHyvK6i+aO0dDQ0NDQOARZvHsx7mBkbpFp/aZh1plTHrOmag2VnkqAZktIhLACQpbICmUAYJalDu6Y1qhNge5UPW00EaKhoaGhoXEIEVbDrKpYxYvrX4xtO6M0PVdMtadlzpI6e8v27KbmbbpGICJCXGHY5g/jDqsd8pHthi7UvPNo7hgNDQ0NDY1DhE93fcqjKx+NWTMAZEmm3lef1vF5lpY5S5wWCMmgUyG7KWL12GDZSrWuHpOcQRhY71UxWxSsrRb6VTKNGPu3i2rdT2iWEA0NDQ0NjUOAT3d9ys2f39xGgACoQuXWJbfy6a5PU+YxNn8sBZYCJCSEJMWsIVFLiJBgbt9PsLQKOvW2c8s4Th9wwBa000SIhoaGhoZGGkTdJB9t/4hVFasIq+EezfvRlY8iSByn8djKx1KWqcgKdxx5R+x7nS3yP8ML+ub5Q06afiYF00tiabzNRSqZRnJ+NRTziNyunUQX0NwxGhoaGhoaKYjnJimwFHDHkXdwUslJ3c6/dUBpPASCCk8Fa6rWML5wfNK8Tio5iT+f8GdmrZhFvb2c6DDdQaEcfnvSHzip5CQaLV54dzsA4WIbuecOxNg/84BZQKJoIuQAEFbDrKlaQ7WnmjxLHmPzx6LISqfTHCrlaGhoaBxIDvZzK+omaW+lqPJUcfPnN/PnE/7cbSHSOqC0J9KdVHISk/tM5pVPJ8MPkWNmDb6NAc31tDmMsbRewQEZjhsPTYTsZ9JRz+kq7GQ3Yk+Wo6GhoXGocLCfW8ncJAKBhMRjKx9jcp/J3RJGrQNKeyIdRFwzeX0GAxERsnfHtww44dTIPr2MOcOA1xnAVefrdH17Ck2E7EfSUc9AWgo72Y2YTh7plgMH/61DQ0NDA3rWAtHV51pPukmSEQ0oTVSWhESBpYCx+WM7lW9eyRBgKQBVuze12WdzGPE6A7gbA+zZWEfxkCzk/zV3zOzZs3niiScoLy9n+PDhPP300xx33HFx077zzjs899xzrFu3Dr/fz/Dhw7nvvvuYPn16m3Rvv/0299xzD9u2baO0tJSHH36Ys88++0CcToxU6hngnqX3oKImTfPg8gdxBpz8sdXUvVEqPZXc9PlNWPXWpHncv+x+ZElOS8kv3rM45VuHJlI0DhQ95WJMleandE3/r7hde9IC0R1rSk+7SRKhyAo3j7uZ27+8vcM+iYgwuP3I2zv9O/YfOI5a/gGAq2x3bPu2tVXUlbtj39//f+uwOowcd8EgSsfkd+UUusRBFSFz587lxhtvZPbs2RxzzDE8//zznHzyyWzcuJG+fft2SP/FF18wdepUHnnkERwOBy+99BKnn346K1asYMyYMQAsX76cCy64gAcffJCzzz6bd999l5kzZ7J06VImTJhwwM4tlXoGcIVcKfOp89fFFSCtic6ol4iGQEPS/VEl/8L3LzB73eyUFpXuuo4ONQ5UPM7/Gt1tk55yMaZKc7DN/e35ubldD7YFojPWlPZ1HZM3hvW169M6z864SRLhD/vjbi+wFHD7kbd36fdz9CmltvmzqKolqAbZ/W0985/veF7uBj/zn1/PjKtGHDAhIgkhEo8H2s9MmDCBsWPH8txzz8W2DR06lLPOOotZs2allcfw4cO54IILuPfeewG44IILcDqdfPzxx7E0M2bMICsri9deey2tPJ1OJ5mZmTQ2NpKRkdGJM2rho+0fxVW0hzI6SUdIhOLuk5DINGTSGGjscDNHVXo6rqMD/RBM9QA8UPE4/2t0t10TdRytrzWI72LsTJpLh1/KnA1zUl7TPUV3rsdU55Jum/SkRbMn7q9Eeby7+V3uXX5vyjo8dtxjnDLglIT1m/729KRiptBSyPxz58e1ApsUE75w6ngJnaRj8czFOEyOlGkTIYTg/Hnn82P9jwDcPeFuMgwZ3X6pUf1+fjx8NAA/9IZBr73F6j814G6IL3gAbFlGfv3wxC67ZjrTh/aYCCkvLycYDMa1YMQjEAhgsVh4880327hKbrjhBtatW8eSJUtS5qGqKv369eP3v/891113HQB9+/blpptu4qabboqle+qpp3j66afZtWtX3Hz8fj9+f8sP4nQ66dOnT7dEyKqKVVy+4PIuHftTpdBSyG3jb+PWJbemfAj2lOWgO+IgnY6uq/E4XXnoH4g26ck8uiMgkrXr78f/nsdXPZ6047DpbKioeEKehGlMiglJkvCGvJ0699b1LbAUMP/c+T1i1erq9RjFrDMnPReDbECSpIRv063Ppyfcrj1xf0FHy2q+OZ9TBpzCe1vfo96fepbQF6e9yLiCcXHrmu5z+LQBp/HB9g9SpkvG+MLxPHfScxgVY+rEcWhd11G5o/j3qf/uVn1a8/34seiavFRlwqbHbyP8Wr+Ux5x10xiKB2d1qbzOiJAec8eceOKJbN68mXA4vclbampqCIfDFBQUtNleUFBARUVFWnn86U9/wu12M3PmzNi2ioqKTuc5a9Ys7r///rTKTJdokFGVpyruQ0VCIt+cDxJJ0ziMjrRuxESkU040XbJJctKhwlPBHV/ekdKHqwq1QyfTFZdOV4N1b/r8Jh6Y+ABPr3k6aSzNI18/QiAciHtO0XwyDBk9EmvTExaZVG3SU51LsjRRAZHONRBPrFZ6KrllyS0djm1POq7MdN5gk9He3N8dcZfKJfDQMQ/x5Oonk96DqcRUQE2+4Fj0fH63+Hd8sfeLDvs743ZNdT5PTnoy5XVw/7L741pWq7xVzNkwJ+m5tObZtc+yz7WPam9LTEaBpYDrx1wfW4E2FakEiMPg4J6j7+nw3Mo15eINeXGH3KyqWMUdX9zBY8c9xrc133b6OvnPpv/EPv9i6C/Sqne6KPl50LSb7CbYUPs9Q+iX8hi3M7GlpCfpMUvIqlWr8Hg8TJo0Ka30ZWVlFBcXs2zZMo4++ujY9ocffph//etf/PDDD0mPf+211/jNb37De++99//bu/e4qOr8f+CvYWC4M4Agg4Vc1FSilIsJmKZlCLqt+nMTFUk317LalctW6nrLe+6aonkpDUS7KOUt8ksJ2WpeyAqlLMsKJYpAwNtAyADD+f3hMjEOlxmYmcPA6+ljHg/mM5/53M445z3nnM/nYPToP76kZTIZdu3ahalTp2rS3n77bcyaNQs1Nc1/KZniSAjwxxcPAK3/aM0dPm0pT+N/5tYClcbTJB2p55n7n8HWr7e2u68dZegpnbZ+NTrYOKC6ruVfyeY0+77ZeOP8Gx0+JWCqX57tKaO1sdeHFazQAN2bZnVGY/3HYsTdI7Ahb0O7gkQAbZ4S6ExcZa6tnnZt/E5qrT9WEis0CB3fvgPcB+D7a98b5UdSR6WNSUNIzxCd7Xvh6gXMyp6lCRLvPGKlz6nZ4qpijD0wFg1CAzztPXFk0hHYSG2M1vbC2X/DrROnAABLX/DFQ1+82OZ7Ou2RkFGjRmH69On4y1/+Arn8jxvcDBli2NQkDw8PSKVSnSMUZWVlOkcy7pSRkYFZs2bhvffe0wpAAEChUBhcpq2tLWxt23cIrTWNq9Y196Xf9CKjtvJYSayQfCxZ5z9i45fC0sjbF652pJ5RPqOw/6f9rR4tMaXGOld9tgq/1/2ORacW6eRpnA00896Z2PfDvlbb2VkCEADYcX5Hs+mN7W8uAGl8va0jB0b55alnGQCw5NQSCP/71xGWEoAAQNblLGRdztJJ13ca/V/6/cViAhCg5QvZG7f5vE/ntXjtWCNjBCAA8ELYC1DWKnXGVeGgQJRfFN688GaHP4sO1g6tntZrVF5dDqmVVOci2Ps870PKyBQ8c/QZNAgNOkes9JlOnPF9hmbMJvefbNQABABsFb3Q2KpbFb9C6q6G+lrLR2ec3Gzh3c/VqG1oicFHQubOnYv33nsPN27cwNixYxEfH4+xY8dCJjP8tr9Dhw5FaGgotm794xd4YGAgxo8f3+KFqXv27MGTTz6JPXv2YMKECTqvx8bGorKyEllZf3xpxMTEwNXV1awXpjZljCl1zX3BKRwUWkFGR+tp7ciNAAFyWzmUKmWrR2TamolD7WNjZYO6hroWXzfGL09j/Xo1NX1PZepzGtJKYgVB6FhQ5WXvhRcfeLHZINFc9O2vo7Ujfq9vfTZdZ9J40Wlz31tny87qdb3HjMAZ2H1hN4DmjwI/O+hZbPlqS5vlpI1Ja3EWjrpBjREZI6CsVTb7emvXF1XXVWP0vtGorK2EjZUNsv+SDQ974967pXzzFlRs3gwAePkvVhgftQ4l77T8w7ujs2MM2YcafAO7TZs2obi4GO+//z6cnZ0xY8YMKBQKPPXUU3pdTNpUcnIy3njjDaSlpeG7775DUlISioqKMGfOHADAggUL8MQTT2jy79mzB0888QReeeUVhIeHo7S0FKWlpbh586YmT0JCArKzs7F27Vp8//33WLt2LT7++GMkJiYa2lWjaYyexwaMxRDFkGbPD7aVZ7TvaByZdARpY9KwdvhapI1Jw0eTPtKKrDtaT+ORm54O2h8+LwcvbBi5AS9FvATgj/+8jRqfL45YrLl7oyVws3Vrsa0SSOBm275DkU3JrAwPzpvTWgACGOeXpzkCEEPHtaXP2vyh8zXX/rQ3jwQSzAic0errU/tPRVuu3LpilACkrc+jq8xV0647XwP06+/Me2d2qI2GaK0/+mqc9trc95a+63IE9ghs8Xtt/cj1mH3/7Fa/tySQQOGgaHWRsLNlZ1sMQADt64vudPjSYVTW3r7FbYx/jNEDEACw9vqj7+5VQIXrL4h+OgiOrtqBiJObrVmn5wLtvIuulZUVoqKikJ6ejitXruD111/H559/jocfftigcmJjY5GSkoLly5dj8ODB+PTTT5GVlQVf39t39yspKUFR0R+Lq7z++uuor6/Hc889B29vb80jISFBkycyMhJ79+7Fzp07cf/99yM9PR0ZGRlmXSPEVPQJMjqqtWCntSBl/cj1iPKLanPH0Bk0fqksCl+keX7n6wCwKHxRm19Ore0YJJBg9n2zTdADy9CRcVU4KPDKQ6+0+FnT5/OoT57ksORWXx/cc7Befe1IAKLv53Fp5NIO97etHa6+2gqY9OmP3FbeoR2/Icuct/a91vSusy21ta1Fwtq7oJkgCFoXpMYNjNOrHEPZNLkcoYdSwPny8+gT3BNPrI7EhKRgPDorEBOSghG/KtKsAQjQwQtTS0tLsXfvXrz11ls4e/YshgwZgjNnzhizfaIw9umYrqa9p45eGPJCmxfZ6jsbyM3WDTdUNzp0sW5L5/GbnubS9+LilsoY5TMKY/aP6dApAVeZq16nuVobE321Na76nAbxcvDSbOuOjKu+U7lNuWKqsafat3Rdl76fR2P0t6OnXRu37/PHn2+2DH37A7R+sXxba7Q0rgHS1mdR3ynW+ox9S/T9nNx5Sif3t1w8lfMUgNszKnfF7GqzjPaouXgRl8dPAAD8934J3p3UE/+d/F9IJKb5cWjSdUKUSiX279+Pd955B8eOHUNAQACmTZuGuLg49O3bt0MN7ywYhHRcW2tIAO2fDaTvlyDQcnDQ0WmvhuwY2upz4+wYU46JMXcugH47D2OMq9j02dHpGzg/N/g57PthX4c/j8ZgjOBA3+1n6Fo+hnwG9A1m9dXesW/rcwLcXtAs6/9lwdvJW5P296N/x/Ffb1/GsO6hdRjjN6bZ93ZU/fXr+DEiEgCQ7y/B6ilSHJl0BL2cepmkPpMGIfb29nBzc8PkyZMRFxdn8KwYS8AgxLSMcfTBmL8a9WGKJcjv7LOpxwQw3s7FWAGEJSx3b6zA+aNJHwFAp+mvMYKDznDLg84SzLb0OWnKy8ELW0dvRR95H3x4+UP86+S/IEDQfD6srUxzJxVBEHBxcDAElQpFHsDzs63xn4f+g2i/aJPUZ9IgJDs7G6NHj4aVVbsuJ7EIDEJMz1i/ki1hJ9aooyummvOXZ3e5SZq+jBUkWhJL2r6dpa3NfU487DzQgAZcq7kGALCV2sLe2h43VDc0eZxsnLBi2AqTfkZ+ihqDuqIiVNkBTyZZY0bgDDw/5HmT1GXyZdvr6+tx7NgxFBQUYNq0aXB2dsZvv/0GFxcXODk5tbvhnQWDkM6hs3yxdCad4Zdnd9UVTi+R6TX3Obmuuo65n8zF+YrzLb5PAolJg9Wfp8ej+ssvAQDx/5Ti3rtDTXYNikmDkJ9//hnR0dEoKiqCSqXCDz/8gICAACQmJqKmpgavvfZahxrfGTAIIaL2YIBHLamqrcLId0fqdW8fU3xmiv/5PJT/938AgISnpLje0x6np52GjZVxF0YDTLxOSEJCAsLCwnD9+nXY29tr0idOnIijR48a3loioi7CHNPoyTJ9d+27FgMQoPW1RIzBusk0XfcqATXqGhTcKDBJXYYw+CqYkydP4tSpUzorpPr6+qK4uNhoDSMiIuoq2ruWiLHYNF2w7PbaaPi6/GsMcB9gkvr0ZfCRkIaGhmbvlPvrr7/C2dnZKI0iIiLqSgxZXM0UrL0Umr8bg5DWrlExF4ODkEcffRQpKSma5xKJBFVVVVi6dCnGjh1rzLYRERF1CSE9Qzq8PHxHNF26vUfV7TZ8U/GNSeoyhMFByIYNG3D8+HEEBgaipqYG06ZNg5+fH4qLi7F27VpTtJGIiMiiGWN5+I5ounR77xpHAEDBjQJU1VaZpD59GRyE9OrVC/n5+Xj++efx9NNPIzg4GC+//DLOnTuHnj3Nu+Y8ERGRpdDnfkemYu3pCfxvmXZF9e0ZMQIEfHv1W5PVqY92Lc9mb2+PJ598Ek8+abx7KhAREXV1o31HY5TPKLNP5ZbY2EDaowfUFRVwvvnHXbnPV5zHUG/xbvBqcBDyySef4MCBAygsLIREIkFAQAAmTZqEESNGmKJ9REREXUrjVG5zs+nZE+qKCtjcqIKkQQLBSoLz5eJenGrQ6Zg5c+Zg9OjR2LNnD65evYry8nK89dZbGDVqFP7xj3+Yqo1ERETUQdaK/82QUTfg7trbq5ufrziPdiycbjR6ByEHDx7Ezp07kZaWhoqKCuTm5uKzzz5DeXk5duzYge3btyMzM9OUbSUiIqJ2ajpDJsTKDwBQfqtc61YD5qZ3ELJz504kJydj5syZkEj+uLLXysoKTz75JBITE5GammqSRhIREVHHNJ0hc2/DH+uGiLleiN5ByNmzZzFx4sQWX580aRLy8vKM0igiIiIyLuuefwQhAbVyzd/v//Q+vij9AuoG3YVITU3vIKSiogJ33XVXi6/fdddduHr1qlEaRURERMbV9HRMbWmp5u/jvx7Hk0eexJj9Y/Dxzx+btU16ByG1tbU694tpytraGrW1tUZpFBERERlX09MxFy6e0Hm9rLoMyceSzRqIGDRFd/HixXBwcGj2terqaqM0iIiIiIxPMzsGgFul7usCBEggwdrP12KUzyiz3AVa7yBkxIgRuHjxYpt5iIiIqPOROjlBsLeD5FYN3Cubn5YrQEBpdSnOlp01y1omegchx44dM2EziIiIyNTqejhD9msNelQCEATNUu53Kq8uN0t7DL53DBEREVkmq56eAAC7OsBe1XI+TwdP87THLLUQERGR6Nzu7qP5272ZG+hKIIHCQYGQniFmaQ+DECIiom5C5vXHxak97rg4VYLbp2bmPTDPLBelAgxCiIiIug1rxR/TdP1UzlqveTl4Yf3I9RjtO9p87TFbTURERCSqpmuFJPjEY8KYISivLoengydCeoaY7QhIIwYhRERE3YR1kyBEXVZmlmm4rTHa6ZiBAwdCKjVvBEVERET6a3r/mPorZSK25DajHQlZs2YNbt68aaziiIiIyMisPXoAUimgVqP+yhWxm2O8IGTChAnGKoqIiIhMQCKVwtrDA/VXrqCuTPwjIZwdQ0RE1I00zpBRX70Koa5O3Lbokyk4OBiSFpZ2vdPZs2c71CAiIiIyHZueXqgBAEFAfXk5bHr1Eq0tegUhTU+11NTUYOvWrQgMDERERAQA4LPPPsO3336LZ5991iSNJCIiIuNoOkOmrvSKqEGIXqdjli5dqnmUl5dj7ty5yM3Nxfr167F+/XqcPn0aiYmJuNKOi1y2bt0Kf39/2NnZITQ0FCdOnGgxb0lJCaZNm4b+/fvDysoKiYmJOnnS09MhkUh0HjU1NQa3jYiIqKux9uqp+bu+TNyLUw2+JuS9997DE088oZM+ffp07N+/36CyMjIykJiYiIULF+LcuXMYPnw4YmJiUFRU1Gx+lUoFT09PLFy4EIMGDWqxXBcXF5SUlGg97OzsDGobERFRV9R0wTKxZ8gYHITY29vj5MmTOuknT540eEe/fv16zJo1C3/7298wcOBApKSkwMfHB9u2bWs2v5+fHzZu3IgnnngCcrm8xXIlEgkUCoXWg4iIiLTXCqkTea0Qg6foJiYm4plnnkFeXh7Cw8MB3L4mJC0tDUuWLNG7nNraWuTl5WH+/Pla6VFRUTh9+rShzdJSVVUFX19fqNVqDB48GCtWrEBwcHCL+VUqFVSqP+5prFQqO1Q/ERFRZ2Wj6DxHQgwOQubPn4+AgABs3LgR77zzDoDbq6Wmp6dj8uTJepdTUVEBtVoNryaHhQDAy8sLpaWlhjZLY8CAAUhPT8d9990HpVKJjRs3YtiwYfjqq6/Qr1+/Zt+zZs0aLFu2rN11EhERWQrrTnQ6pl2LlU2ePNmggKM1d079FQRB7+nAzQkPD9ccoQGAYcOGISQkBK+++io2bdrU7HsWLFiA5ORkzXOlUgkfH592t4GIiKizsrK3h5WLCxqUStRZ2jUhAQEBuHr1qk76jRs3EBAQoHc5Hh4ekEqlOkc9ysrKdI6OdISVlRWGDBmCH3/8scU8tra2cHFx0XoQERF1VTb/myFTf+UKBEEQrR0GByGFhYVQq9U66SqVCsXFxXqXI5PJEBoaipycHK30nJwcREZGGtqsFgmCgPz8fHh7exutTCIiIkvWeHGqUFsL9Y0b4rVD34yZmZmav48cOaI1O0WtVuPo0aPw8/MzqPLk5GTEx8cjLCwMERER2L59O4qKijBnzhwAt0+TFBcXY/fu3Zr35OfnA7h98Wl5eTny8/Mhk8kQGBgIAFi2bBnCw8PRr18/KJVKbNq0Cfn5+diyZYtBbSMiIuqqtK4LKSuDtZubOO3QN2PjqqkSiQQzZszQes3GxgZ+fn545ZVXDKo8NjYWV69exfLly1FSUoKgoCBkZWXB19cXwO3Fye5cM6TpLJe8vDy888478PX1RWFhIYDbp4WeeuoplJaWQi6XIzg4GJ9++ikeeOABg9pGRETUVenMkOnfX5R2SAQDTwb5+/vjiy++gIeHh6naJDqlUgm5XI6bN2/y+hAiIupyru/NQOlLLwEAFCuWw+3xx41WtiH7UINnx1y+fLndDSMiIiLxaS3dXireDBm9L0w9c+YMPvzwQ6203bt3w9/fHz179sRTTz2lteAXERERdU5aS7eLeP8YvYOQl156CV9//bXm+fnz5zFr1iyMHj0a8+fPxwcffIA1a9aYpJFERERkPFp30hVxrRC9g5D8/Hw88sgjmud79+7F0KFDsWPHDiQnJ2PTpk149913TdJIIiIiMh6pmxtgYwMAqBfx/jF6ByHXr1/XWkTs+PHjiI6O1jwfMmQIfvnlF+O2joiIiIxOYmUFm55/LFgmFr2DEC8vL81FqbW1tTh79iwiIiI0r1dWVsLmf1EVERERdW7W/wtC1DduoOrESQjNLERqanoHIdHR0Zg/fz5OnDiBBQsWwMHBAcOHD9e8/vXXX6NPnz4maSQREREZjzI7GzXffad5/svs2fjpkdFQZmebtR16ByErV66EVCrFQw89hB07dmDHjh2QyWSa19PS0hAVFWWSRhIREZFxKLOzUZyQCKGmRiu9/soVFCckmjUQMXixsps3b8LJyQlSqVQr/dq1a3ByctIKTCwVFysjIqKuSFCr8dMjo1F/x81jNSQSWHt5oe/RjyG5Yz+vL0P2oQbfwE4ul+sEIADg7u7eJQIQIiKirqr6y7yWAxAAEATUl5ai+ss8s7TH4CCEiIiILFN9eblR83UUgxAiIqJuwtrT06j5OopBCBERUTfhEBYKa4UCkEiazyCRwFqhgENYqFnawyCEiIiom5BIpfD614L/PbkjEPnfc69/LWj3RamGYhBCRETUjbhEReGujSla948Bbt9P5q6NKXAx43Ib1mariYiIiDoFl6goOD/yyO3ZMuXlsPb0hENYqNmOgDRiEEJERNQNSaRSOA59QNQ28HQMERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBCBEREYmCQQgRERGJQvQgZOvWrfD394ednR1CQ0Nx4sSJFvOWlJRg2rRp6N+/P6ysrJCYmNhsvv379yMwMBC2trYIDAzEwYMHTdR6IiIiai9Rg5CMjAwkJiZi4cKFOHfuHIYPH46YmBgUFRU1m1+lUsHT0xMLFy7EoEGDms2Tm5uL2NhYxMfH46uvvkJ8fDwmT56MM2fOmLIrREREZCCJIAiCWJUPHToUISEh2LZtmyZt4MCBmDBhAtasWdPqe0eOHInBgwcjJSVFKz02NhZKpRIffvihJi06Ohpubm7Ys2ePXu1SKpWQy+W4efMmXFxc9O8QERFRN2fIPlS0IyG1tbXIy8tDVFSUVnpUVBROnz7d7nJzc3N1yhwzZkyrZapUKiiVSq0HERERmZZoQUhFRQXUajW8vLy00r28vFBaWtrucktLSw0uc82aNZDL5ZqHj49Pu+snIiIi/Yh+YapEItF6LgiCTpqpy1ywYAFu3rypefzyyy8dqp+IiIjaZi1WxR4eHpBKpTpHKMrKynSOZBhCoVAYXKatrS1sbW3bXScREREZTrQjITKZDKGhocjJydFKz8nJQWRkZLvLjYiI0CkzOzu7Q2USERGR8Yl2JAQAkpOTER8fj7CwMERERGD79u0oKirCnDlzANw+TVJcXIzdu3dr3pOfnw8AqKqqQnl5OfLz8yGTyRAYGAgASEhIwIgRI7B27VqMHz8e77//Pj7++GOcPHnS7P0jIiKilokahMTGxuLq1atYvnw5SkpKEBQUhKysLPj6+gK4vTjZnWuGBAcHa/7Oy8vDO++8A19fXxQWFgIAIiMjsXfvXixatAiLFy9Gnz59kJGRgaFDh5qtX0RERNQ2UdcJ6ay4TggREVH7WMQ6IURERNS9MQghIiIiUTAIISIiIlEwCCEiIiJRMAghIiIiUTAIISIiIlEwCCEiIiJRMAghIiIiUTAIISIiIlGIumy7pVOr1airqxO7GUREXYJMJoOVFX8bdycMQtpBEASUlpbixo0bYjeFiKjLsLKygr+/P2QymdhNITNhENIOjQFIz5494eDgAIlEInaTiIgsWkNDA3777TeUlJSgd+/e/F7tJhiEGEitVmsCkB49eojdHCKiLsPT0xO//fYb6uvrYWNjI3ZzyAx48s1AjdeAODg4iNwSIqKupfE0jFqtFrklZC4MQtqJhwqJiIyL36vdD4MQIiIiEgWDECIiIhIFg5BuQiKRtPqYOXOmJt+hQ4d03j9z5kxMmDBB63lz5URHR+PYsWNt1peeng4AuHXrFtzc3ODu7o5bt27p1Lt9+3aMHDkSLi4ukEgknBbdAkvdvo0EQUBMTEyL7evuLHH7FhYWtvj+9957z0QjRZaGs2O6iZKSEs3fGRkZWLJkCS5evKhJs7e3N7jM6Oho7Ny5UyvN1tYWjo6OWvUlJCRAqVRq5ZXL5QCA/fv3IygoCIIg4MCBA4iLi9Mqr7q6GtHR0YiOjsaCBQsMbmN3Yanbt1FKSgqvB2iFJW5fHx8frXKA2z8q/v3vfyMmJsbg9lLXxCCkm1AoFJq/5XI5JBKJVlp72NratlhG03R7e3uoVKpm86ampmL69OkQBAGpqak6O6nExEQAwLFjxzrU1q7OUrcvAHz11VdYv349vvjiC3h7e3eozV2VJW5fqVSq856DBw8iNjYWTk5OHWo7dR0MQozksVdPorxSZfZ6PZ1t8cE/HjR7vcZQUFCA3NxcHDhwAIIgIDExEZcuXUJAQIDYTdNxedJfUF9RYfZ6rT084L9/n9nrNQZ9tm91dTWmTp2KzZs3d3in2hFXXj2Hhspas9dr5SyD1z+CzV6vMRj6/zcvLw/5+fnYsmWLmVtKnRmDECMpr1ShVFkjdjOMYurUqZBKpVppKpUK48aN00o7fPiwzi+aefPmYfHixXrVk5aWhpiYGLi5uQG4fXg4LS0NK1eu7EDrTaO+ogL1V66I3Qyj6EzbNykpCZGRkRg/fnx7umI0DZW1UCvNH4SYQmfavk2lpqZi4MCBiIyM1Lcr1A0wCDEST2fbLlPvhg0bMHr0aK20efPm6SwgNGrUKGzbtk0rzd3dXa861Go1du3ahY0bN2rSpk+fjqSkJCxbtkznS1Rs1h4eXabezrJ9MzMz8cknn+DcuXPt7InxWDmLc68SU9TbWbZvU7du3cI777yjd4BD3QeDECOx1FMizVEoFOjbt69WmrOzs87MFEdHR518+jpy5AiKi4sRGxurla5Wq5Gdnd3pLlyz1FMizeks2/eTTz5BQUEBXF1dtfJMmjQJw4cPN+t1QJZ6SqQ5nWX7NrVv3z5UV1fjiSeeaFd91HVxii6JIjU1FVOmTEF+fr7WIy4uDqmpqWI3jzpIn+07f/58fP3111qvA7d/yd85a4M6F0P//6ampuLPf/4zPD09RWgtdWY8EkLtplKpUFpaqpVmbW0NjzZOIZSXl+ODDz5AZmYmgoKCtF6bMWMGxo0bh/Lycnh6eqK0tBSlpaX46aefAADnz5+Hs7MzevfurfehY2ofU29fhULR7MWovXv3hr+/f8c7QK0yx/9fAPjpp5/w6aefIisry7gdoC6BR0Ko3T766CN4e3trPR58sO3TUrt374ajoyMeeeQRnddGjRoFZ2dnvPnmmwCA1157DcHBwZg9ezYAYMSIEQgODkZmZqZxO0M6zLF9STzm2r5paWm46667EBUVZdT2U9cgEQRBELsRnY1SqYRcLsfNmzfh4uKi9VpNTQ0uX74Mf39/2NnZidRCIqKuh9+vXUNr+9A78UgIERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBCBEREYmCQQgRERGJQvQgZOvWrZrpWKGhoThx4kSr+Y8fP47Q0FDY2dkhICAAr732mtbr6enpkEgkOo+amq5xczkiIqKuQtQgJCMjA4mJiVi4cCHOnTuH4cOHIyYmBkVFRc3mv3z5MsaOHYvhw4fj3Llz+Ne//oW5c+di//79WvlcXFxQUlKi9eCccyIios5F1GXb169fj1mzZuFvf/sbACAlJQVHjhzBtm3bsGbNGp38r732Gnr37o2UlBQAwMCBA/Hll19i3bp1mDRpkiafRCJpdjloIiIi6jxEOxJSW1uLvLw8naV8o6KicPr06Wbfk5ubq5N/zJgx+PLLL1FXV6dJq6qqgq+vL+6++2786U9/avNW4SqVCkqlUutBREREpiVaEFJRUQG1Wg0vLy+tdC8vL52bKjUqLS1tNn99fT0qKioAAAMGDEB6ejoyMzOxZ88e2NnZYdiwYfjxxx9bbMuaNWsgl8s1Dx8fnw72rvNp7jqZpo+ZM2dq8h06dEjn/TNnzsSECRO0njdXTnR0NI4dO9Zmfenp6QCAW7duwc3NDe7u7rh165ZWndeuXcM//vEP9O/fHw4ODujduzfmzp2LmzdvmmiULJclbl8AGDlypM57p0yZYoIRsmyWun1LS0sRHx8PhUIBR0dHhISEYN++fSYYIbJUot9FVyKRaD0XBEEnra38TdPDw8MRHh6ueX3YsGEICQnBq6++ik2bNjVb5oIFC5CcnKx5rlQqu1wgUlJSovk7IyMDS5YswcWLFzVp9vb2BpcZHR2tc8t1W1tbODo6atWXkJAApVKplVculwMA9u/fj6CgIAiCgAMHDiAuLk6T57fffsNvv/2GdevWITAwED///DPmzJmD3377jV9kd7DE7dto9uzZWL58eYfa2tVZ6vaNj4/HzZs3kZmZCQ8PD7zzzjuIjY3Fl19+ieDgYIPbTF2PaEGIh4cHpFKpzlGPsrIynaMdjRQKRbP5ra2t0aNHj2bfY2VlhSFDhrR6JMTW1ha2trYG9sCyNL1GRi6XG+W6GVtb2xbLaJpub28PlUrVbN7U1FRMnz4dgiAgNTVV60ssKChI66LjPn36YNWqVZg+fTrq6+thbS16DN1pWOL2beTg4MBruNpgqds3NzcX27ZtwwMPPAAAWLRoETZs2ICzZ88yCCEAIgYhMpkMoaGhyMnJwcSJEzXpOTk5GD9+fLPviYiIwAcffKCVlp2djbCwMNjY2DT7HkEQkJ+fj/vuu894jW/O6w8BVWWmraM5Tj2Bp4+bv14jKCgoQG5uLg4cOABBEJCYmIhLly4hICCgxfc03pXR3AHIu6u/QLWy1qx1AoCDiwyT/zXE7PUag77b9+2338Zbb70FLy8vxMTEYOnSpXB2djZrW19//XVUVVWZtU4AcHJywtNPP232eo1Bn+374IMPIiMjA+PGjYOrqyveffddqFQqjBw5UryGU6ci6k/J5ORkxMfHIywsDBEREdi+fTuKioowZ84cALdPkxQXF2P37t0AgDlz5mDz5s1ITk7G7NmzkZubi9TUVOzZs0dT5rJlyxAeHo5+/fpBqVRi06ZNyM/Px5YtW0zbmaoyoPI309ZhJlOnToVUKtVKU6lUGDdunFba4cOH4eTkpJU2b948LF68WK960tLSEBMTAzc3NwC3Dw+npaVh5cqVzea/evUqVqxYIcqXdrWyFr/fUJm9XlPoTNs3Li4O/v7+UCgU+Oabb7BgwQJ89dVXyMnJaU/X2q2qqgqVlZVmrdNUOtP2zcjIQGxsLHr06AFra2s4ODjg4MGD6NOnT3u6Rl2QqEFIbGwsrl69iuXLl6OkpARBQUHIysqCr68vgNvnQZuuGeLv74+srCwkJSVhy5Yt6NWrFzZt2qQ1PffGjRt46qmnUFpaCrlcjuDgYHz66aeaw4Em49TTtOWbsd4NGzZg9OjRWmnz5s2DWq3WShs1ahS2bdumlebu7q5XHWq1Grt27cLGjRs1adOnT0dSUhKWLVum8yWqVCoxbtw4BAYGYunSpYZ0xygcXGRmr9NU9Xam7Tt79mzN60FBQejXrx/CwsJw9uxZhISEGNSvjrhzZ2zJ9Xam7bto0SJcv34dH3/8MTw8PHDo0CE8/vjjOHHihOmPTpNFEP2k+rPPPotnn3222dcar8Bu6qGHHsLZs2dbLG/Dhg3YsGGDsZqnPws9JdIchUKBvn37aqU5Ozvjxo0bWmmOjo46+fR15MgRFBcXIzY2VitdrVYjOzsbMTExmrTKykpER0fDyckJBw8ebPHUmylZ6imR5nS27dtUSEgIbGxs8OOPP5o1CLHUUyLN6Szbt6CgAJs3b8Y333yDe++9FwAwaNAgnDhxAlu2bNFZ7Zq6J9GXbafuKTU1FVOmTEF+fr7WIy4uDqmpqZp8SqUSUVFRkMlkyMzM5Mq3FkLf7Xunb7/9FnV1dfD29jZja8lQ+mzf6upqALcnBzQllUrR0NBg9jZT5yT6kRCyXCqVSme2krW1NTw8PFp9X3l5OT744ANkZmYiKChI67UZM2Zg3LhxKC8vh52dHaKiolBdXY233npLayE5T09PnVM2ZFym3r5KpRJvv/02xo4dCw8PD1y4cAH//Oc/ERwcjGHDhhm9P6TN1Nt3wIAB6Nu3L55++mmsW7cOPXr0wKFDh5CTk4PDhw8bvT9kmRiEULt99NFHOr9Y+/fvj++//77V9+3evRuOjo545JFHdF4bNWoUnJ2d8eabbyIkJARnzpwBAJ3DxpcvX4afn1/HOkCtMvX2ffzxx3H06FFs3LgRVVVV8PHxwbhx47B06VIGmGZg6u2bnJyMrKwszJ8/H4899hiqqqrQt29f7Nq1C2PHjjVqX8hySYTG1b5IQ6lUQi6Xa6aDNlVTU4PLly9r7vxLRETGwe/XrqG1feideE0IERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBCBEREYmCQQgRERGJgkEIERERiYJBSDchkUhafcycOVOT79ChQzrvnzlzJiZMmKD1vLlyoqOjcezYsTbrS09PBwDcunULbm5ucHd3x61bt3Tqffrpp9GnTx/Y29vD09MT48ePb/PeFt2RpW5fAMjNzcXDDz8MR0dHuLq6YuTIkS3m7a4sdfsWFBRg4sSJ8PT0hIuLCyZPnowrV66YYITIUvEGdt1ESUmJ5u+MjAwsWbIEFy9e1KTZ29sbXGZ0dDR27typlWZrawtHR0et+hISEqBUKrXyyuVyAMD+/fsRFBQEQRBw4MABxMXFaZUXGhqKuLg49O7dG9euXcNLL72EqKgoXL58mTc5a8JSt29ubi6io6OxYMECvPrqq5DJZPjqq690bv/e3Vni9v39998RFRWFQYMG4ZNPPgEALF68GI899hg+++wzbmMCwCCk21AoFJq/5XI5JBKJVlp72NratlhG03R7e3uoVKpm86ampmL69OkQBAGpqak6O6mnnnpK87efnx9WrlyJQYMGobCwEH369OlQ+7sSS92+SUlJmDt3LubPn69J69evX4fa3RVZ4vY9deoUCgsLce7cOc1NzHbu3Al3d3d88sknGD16dIfaT10DgxAjiT0ci4pbFWav18PeAxl/yjB7vcZQUFCA3NxcHDhwAIIgIDExEZcuXUJAQECz+X///Xfs3LkT/v7+8PHxMWtb31qQiN9vXDdrnQDg6OqG6WtSzF6vMbS1fcvKynDmzBnExcUhMjISBQUFGDBgAFatWoUHH3zQrG39/IvxqK01//9fmcwDDwx53+z1GkNb21elUkEikcDW1lbzHjs7O1hZWeHkyZMMQggAgxCjqbhVgbLqMrGbYRRTp07VOdWhUqkwbtw4rbTDhw/DyclJK23evHlYvHixXvWkpaUhJiYGbm5uAG4fHk5LS8PKlSu18m3duhUvvvgifv/9dwwYMAA5OTmQyWSGdqtDfr9xHVXXrpq1TlPpLNv30qVLAICXXnoJ69atw+DBg7F792488sgj+Oabb8x6RKS2tgIqVanZ6jOlzrJ9w8PD4ejoiHnz5mH16tUQBAHz5s1DQ0OD1uke6t4YhBiJh71Hl6l3w4YNOr9S5s2bB7VarZU2atQobNu2TSvN3d1drzrUajV27dqFjRs3atKmT5+OpKQkLFu2TOtLNC4uDo8++ihKSkqwbt06TJ48GadOnYKdnZ2hXWs3R1c3s9Vl6no7y/ZtaGgAcPvi47/+9a8AgODgYBw9ehRpaWlYs2aNwX1rL5lMnP+/pqi3s2xfT09PvPfee3jmmWewadMmWFlZYerUqQgJCeH1XKTBIMRILPWUSHMUCgX69u2rlebs7IwbN25opTk6Ourk09eRI0dQXFyM2NhYrXS1Wo3s7GzExMRo0uRyOeRyOfr164fw8HC4ubnh4MGDmDp1arvqbg9LPSXSnM6yfb29vQEAgYGBWnkGDhyIoqKidtXbXpZ6SqQ5nWX7AkBUVBQKCgpQUVEBa2truLq6QqFQwN/fv131UtfDy5NJFKmpqZgyZQry8/O1HnFxcUhNTW31vYIgQKVSmaml1B76bF8/Pz/06tVLa5YHAPzwww/w9fUVo9mkJ0P//3p4eMDV1RWffPIJysrK8Oc//1mEVlNnxCMh1G4qlQqlpdrn0a2treHh0foh5vLycnzwwQfIzMxEUFCQ1mszZszAuHHjUF5ejsrKSmRkZCAqKgqenp4oLi7G2rVrYW9vj7Fjxxq9P6TN1NvX09MTL7zwApYuXYpBgwZh8ODB2LVrF77//nvs27fP6P0hbebYvjt37sTAgQPh6emJ3NxcJCQkICkpCf379zd6f8gyMQihdvvoo480h9Qb9e/fv83FxHbv3g1HR0c88sgjOq+NGjUKzs7OePPNNzFlyhScOHECKSkpuH79Ory8vDBixAicPn0aPXv2NGpfSJept29ycjISExNRU1ODpKQkXLt2DYMGDUJOTg6nX5uBObbvxYsXsWDBAly7dg1+fn5YuHAhkpKSjNoPsmwSQRAEsRvR2SiVSsjlcty8eVMzv71RTU0NLl++DH9/f7NeGElE1NXx+7VraG0feideE0JERESiYBBCREREomAQQkRERKJgEEJERESiYBDSTryel4jIuPi92v0wCDGQjY0NAKC6ulrklhARdS21tbUAwGXduxGuE2IgqVQKV1dXlJXdvlmdg4MDJBKJyK0iIrJsDQ0NKC8vh4ODA6ytuWvqLril20GhUACAJhAhIqKOs7KyQu/evfnDrhthENIOEokE3t7e6NmzJ+rq6trMr24Q8PWvN3Dt91q4O8pw/92ukFpJjJ7HkuqxpLZyTDgmxsxDLZPJZLCy4lUC3QmDkA6QSqVtnrv86JsSLPvgAkpu1mjSvOV2WPpYIKKDvI2Wx5LqsaS2ckw4JsaqB7gdpHx++RrKKmvQ09kOD/i7NxvItJbHGGV0tXosqa2WVo+pib5s+9atW/Gf//wHJSUluPfee5GSkoLhw4e3mP/48eNITk7Gt99+i169euHFF1/EnDlztPLs378fixcvRkFBAfr06YNVq1Zh4sSJerfJkCVnW/PRNyV45q2zuHOAGzfxtukhANDhPE+N8Mf2Ty9bRD2W1FaOiXj1WFJbDcljKQGTJdVjSW21tHray5B9qKhBSEZGBuLj47F161YMGzYMr7/+Ot544w1cuHABvXv31sl/+fJlBAUFYfbs2Xj66adx6tQpPPvss9izZw8mTZoEAMjNzcXw4cOxYsUKTJw4EQcPHsSSJUtw8uRJDB06VK92GSMIUTcIeHDtJ1ob+E5ezjIAElypbPm29PrkkQA6X36dtZ6OliEB4OViC0CCUmXLY2slARpaqEifMsxVjyW1lWPS/jxyBxvcrK6ziIDJkuqxpLZaWj0dCUQsJggZOnQoQkJCsG3bNk3awIEDMWHCBKxZs0Yn/7x585CZmYnvvvtOkzZnzhx89dVXyM3NBQDExsZCqVTiww8/1OSJjo6Gm5sb9uzZo1e7jBGE5BZcxdQdn7XrvUTUvTQeAW8pIDJWGcaoRyoBAAnUrew69MljjHraLON/HVa30mHr/+Wp78Cg6FOGMeqxuT0oqFO3XIZMKgEgQa26oV11SAAo5HY4Oe/hdp+aMWQfKto1IbW1tcjLy8P8+fO10qOionD69Olm35Obm4uoqCittDFjxiA1NRV1dXWwsbFBbm6uzq2ix4wZg5SUlBbbolKpoFL98cv75s2bAG4PZHsVlpSjQcW1RIiobe3bXRheRlerxxhl1BqhHn3KMEY9LR9j/kPLx/30V1xWjf9+/TMeCHBv1/sb9536HOMQLQipqKiAWq2Gl5eXVrqXlxdKS0ubfU9paWmz+evr61FRUQFvb+8W87RUJgCsWbMGy5Yt00n38fHRtztERERdxqMpHS+jsrIScrm81Tyiz465cz64IAitzhFvLv+d6YaWuWDBAiQnJ2ueNzQ04Nq1a+jRo4fR5qsrlUr4+Pjgl19+6dDFrqSN42p8HFPT4LgaH8fU+IwxpoIgoLKyEr169Wozr2hBiIeHB6RSqc4RirKyMp0jGY0UCkWz+a2trdGjR49W87RUJgDY2trC1tZWK83V1VXfrhjExcWF/1lMgONqfBxT0+C4Gh/H1Pg6OqZtHQFpJNqqMDKZDKGhocjJydFKz8nJQWRkZLPviYiI0MmfnZ2NsLAwzT1dWsrTUplEREQkDlFPxyQnJyM+Ph5hYWGIiIjA9u3bUVRUpFn3Y8GCBSguLsbu3bsB3J4Js3nzZiQnJ2P27NnIzc1Famqq1qyXhIQEjBgxAmvXrsX48ePx/vvv4+OPP8bJkydF6SMRERE1T9QgJDY2FlevXsXy5ctRUlKCoKAgZGVlwdfXFwBQUlKCoqIiTX5/f39kZWUhKSkJW7ZsQa9evbBp0ybNGiEAEBkZib1792LRokVYvHgx+vTpg4yMDL3XCDEVW1tbLF26VOe0D3UMx9X4OKamwXE1Po6p8Zl7TEVfMZWIiIi6J94piIiIiETBIISIiIhEwSCEiIiIRMEghIiIiETBIKQDtm7dCn9/f9jZ2SE0NBQnTpxoNf+WLVswcOBA2Nvbo3///pqpx03t378fgYGBsLW1RWBgIA4ePGiq5ndKxh7THTt2YPjw4XBzc4ObmxtGjx6Nzz//3JRd6JRM8VlttHfvXkgkEkyYMMHIre7cTDGmN27cwHPPPQdvb2/Y2dlh4MCByMrKMlUXOh1TjGlKSgr69+8Pe3t7+Pj4ICkpCTU1xrjDSuf36aef4rHHHkOvXr0gkUhw6NChNt9z/PhxhIaGws7ODgEBAXjttdd08hh1PyVQu+zdu1ewsbERduzYIVy4cEFISEgQHB0dhZ9//rnZ/Fu3bhWcnZ2FvXv3CgUFBcKePXsEJycnITMzU5Pn9OnTglQqFVavXi189913wurVqwVra2vhs88+M1e3RGWKMZ02bZqwZcsW4dy5c8J3330n/PWvfxXkcrnw66+/mqtbojPFuDYqLCwU7rrrLmH48OHC+PHjTdyTzsMUY6pSqYSwsDBh7NixwsmTJ4XCwkLhxIkTQn5+vrm6JSpTjOlbb70l2NraCm+//bZw+fJl4ciRI4K3t7eQmJhorm6JKisrS1i4cKGwf/9+AYBw8ODBVvNfunRJcHBwEBISEoQLFy4IO3bsEGxsbIR9+/Zp8hh7P8UgpJ0eeOABYc6cOVppAwYMEObPn99s/oiICOH555/XSktISBCGDRumeT558mQhOjpaK8+YMWOEKVOmGKnVnZspxvRO9fX1grOzs7Br166ON9hCmGpc6+vrhWHDhglvvPGGMGPGjG4VhJhiTLdt2yYEBAQItbW1xm+wBTDFmD733HPCww8/rJUnOTlZePDBB43UasuhTxDy4osvCgMGDNBKe/rpp4Xw8HDNc2Pvp3g6ph1qa2uRl5eHqKgorfSoqCicPn262feoVCrY2dlppdnb2+Pzzz9HXV0dACA3N1enzDFjxrRYZldiqjG9U3V1Nerq6uDu3r5bVFsaU47r8uXL4enpiVmzZhm/4Z2YqcY0MzMTEREReO655+Dl5YWgoCCsXr0aarXaNB3pREw1pg8++CDy8vI0p2AvXbqErKwsjBs3zgS9sHwt7YO+/PJLk+2nGIS0Q0VFBdRqtc5N8by8vHRuntdozJgxeOONN5CXlwdBEPDll18iLS0NdXV1qKioAACUlpYaVGZXYqoxvdP8+fNx1113YfTo0UbvQ2dkqnE9deoUUlNTsWPHDpP3obMx1ZheunQJ+/btg1qtRlZWFhYtWoRXXnkFq1atMnmfxGaqMZ0yZQpWrFiBBx98EDY2NujTpw9GjRqF+fPnm7xPlqilfVB9fb3J9lMMQjpAIpFoPRcEQSet0eLFixETE4Pw8HDY2Nhg/PjxmDlzJgBAKpW2q8yuyBRj2ujf//439uzZgwMHDuj8gurqjDmulZWVmD59Onbs2AEPDw9TN73TMvZntaGhAT179sT27dsRGhqKKVOmYOHChdi2bZtJ+9GZGHtMjx07hlWrVmHr1q04e/YsDhw4gMOHD2PFihUm7Ycla24b3JluzP0Ug5B28PDwgFQq1Yn8ysrKdCLERvb29khLS0N1dTUKCwtRVFQEPz8/ODs7a77IFQqFQWV2JaYa00br1q3D6tWrkZ2djfvvv99k/ehsTDGuBQUFKCwsxGOPPQZra2tYW1tj9+7dyMzMhLW1NQoKCszRNdGY6rPq7e2Ne+65RyuAHjhwIEpLS1FbW2u6DnUCphrTxYsXIz4+Hn/7299w3333YeLEiVi9ejXWrFmDhoYGk/fL0rS0D7K2tkaPHj1azdPe/RSDkHaQyWQIDQ1FTk6OVnpOTg4iIyNbfa+NjQ3uvvtuSKVS7N27F3/6059gZXV7M0REROiUmZ2d3WaZXYGpxhQA/vOf/2DFihX46KOPEBYWZpL2d1amGNcBAwbg/PnzyM/P1zz+/Oc/Y9SoUcjPz4ePj48puyQ6U31Whw0bhp9++klr5/jDDz/A29sbMpnM+B3pREw1ptXV1VrfBcDtoyTC7UkZxu1EF9DSPigsLAw2Njat5mn3fqpdl7OSZjpZamqqcOHCBSExMVFwdHQUCgsLBUEQhPnz5wvx8fGa/BcvXhTefPNN4YcffhDOnDkjxMbGCu7u7sLly5c1eU6dOiVIpVLh5ZdfFr777jvh5Zdf7pZTdI05pmvXrhVkMpmwb98+oaSkRPOorKw0d/dEY4pxvVN3mx1jijEtKioSnJychL///e/CxYsXhcOHDws9e/YUVq5cae7uicIUY7p06VLB2dlZ2LNnj3Dp0iUhOztb6NOnjzB58mRzd08UlZWVwrlz54Rz584JAIT169cL586d00x7vnNMG6foJiUlCRcuXBBSU1N1pugaez/FIKQDtmzZIvj6+goymUwICQkRjh8/rnltxowZwkMPPaR5fuHCBWHw4MGCvb294OLiIowfP174/vvvdcp87733hP79+ws2NjbCgAEDhP3795ujK52GscfU19dXAKDzWLp0qZl61DmY4rPaVHcLQgTBNGN6+vRpYejQoYKtra0QEBAgrFq1SqivrzdHdzoFY49pXV2d8NJLLwl9+vQR7OzsBB8fH+HZZ58Vrl+/bqYeieu///1vs99/M2bMEARBd0wFQRCOHTsmBAcHCzKZTPDz8xO2bdumU64x91MSQeAxKSIiIjI/XhNCREREomAQQkRERKJgEEJERESiYBBCREREomAQQkRERKJgEEJERESiYBBCREREomAQQkRERKJgEEJE3caxY8cgkUhw48YNAEB6ejpcXV1FbRNRd8YghIiM5pdffsGsWbPQq1cvyGQy+Pr6IiEhAVevXjV7W0aOHInExESttMjISJSUlEAul5u9PUSki0EIERnFpUuXEBYWhh9++AF79uzBTz/9hNdeew1Hjx5FREQErl27JnYTIZPJoFAoIJFIxG4KEYFBCBEZyXPPPQeZTIbs7Gw89NBD6N27N2JiYvDxxx+juLgYCxcuBABIJBIcOnRI672urq5IT0/XPJ83bx7uueceODg4ICAgAIsXL0ZdXZ3m9ZdeegmDBw/Gm2++CT8/P8jlckyZMgWVlZUAgJkzZ+L48ePYuHEjJBIJJBIJCgsLdU7HNOeDDz5AaGgo7OzsEBAQgGXLlqG+vt5o40REf2AQQkQddu3aNRw5cgTPPvss7O3ttV5TKBSIi4tDRkYG9L1fprOzM9LT03HhwgVs3LgRO3bswIYNG7TyFBQU4NChQzh8+DAOHz6M48eP4+WXXwYAbNy4EREREZg9ezZKSkpQUlICHx+fNus9cuQIpk+fjrlz5+LChQt4/fXXkZ6ejlWrVuk5EkRkCAYhRNRhP/74IwRBwMCBA5t9feDAgbh+/TrKy8v1Km/RokWIjIyEn58fHnvsMfzzn//Eu+++q5WnoaEB6enpCAoKwvDhwxEfH4+jR48CAORyOWQyGRwcHKBQKKBQKCCVStusd9WqVZg/fz5mzJiBgIAAPProo1ixYgVef/11vdpNRIaxFrsBRNT1NR4BkclkeuXft28fUlJS8NNPP6Gqqgr19fVwcXHRyuPn5wdnZ2fNc29vb5SVlXWonXl5efjiiy+0jnyo1WrU1NSguroaDg4OHSqfiLTxSAgRdVjfvn0hkUhw4cKFZl///vvv4enpCVdXV0gkEp3TMk2v9/jss88wZcoUxMTE4PDhwzh37hwWLlyI2tparffY2NhoPZdIJGhoaOhQPxoaGrBs2TLk5+drHufPn8ePP/4IOzu7DpVNRLp4JISIOqxHjx549NFHsXXrViQlJWldF1JaWoq3334bzz33HADA09MTJSUlmtd//PFHVFdXa56fOnUKvr6+mgtZAeDnn382uE0ymQxqtdqg94SEhODixYvo27evwfURkeEYhBCRUWzevBmRkZEYM2YMVq5cCX9/f3z77bd44YUXcM8992DJkiUAgIcffhibN29GeHg4GhoaMG/ePK2jGn379kVRURH27t2LIUOG4P/+7/9w8OBBg9vj5+eHM2fOoLCwEE5OTnB3d2/zPUuWLMGf/vQn+Pj44PHHH4eVlRW+/vprnD9/HitXrjS4DUTUOp6OISKj6NevH7744gsEBARg8uTJ8PX1RUxMDO655x6cOnUKTk5OAIBXXnkFPj4+GDFiBKZNm4bnn39e61qL8ePHIykpCX//+98xePBgnD59GosXLza4Pc8//zykUikCAwPh6emJoqKiNt8zZswYHD58GDk5ORgyZAjCw8Oxfv16+Pr6Glw/EbVNIug7Z46IyEBLly7F+vXrkZ2djYiICLGbQ0SdDIMQIjKpnTt34ubNm5g7dy6srHjwlYj+wCCEiIiIRMGfJURERCQKBiFEREQkCgYhREREJAoGIURERCQKBiFEREQkCgYhREREJAoGIURERCQKBiFEREQkCgYhREREJIr/D/maDmkGBMLjAAAAAElFTkSuQmCC", + "text/plain": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "### Variability in Theta by Integrated Posterior Quantile using Posterior Draws\n", + "mat_sort = mat[np.argsort(POST_pool),:]\n", + "mat_trans_sort = mat_trans[np.argsort(POST_pool),:]\n", + "POST_sort = np.sort(POST_pool)\n", + "QUANTS = np.linspace(0.9,0.999,50)\n", + "THETA_VARS= np.empty((0,mat_sort.shape[1]))\n", + "for i in range(len(QUANTS)):\n", + " THETA_VARS = np.append(THETA_VARS,mat_trans_sort[np.where(POST_sort >= np.quantile(POST_sort,QUANTS[i])),:].var(axis=1),axis=0)\n", + "\n", + "KEYS = list(['THETA1','THETA2','THETA3','THETA4','THETA5','THETA6','THETA7','THETA8','THETA9'])\n", + "import seaborn as sns\n", + "fig,ax=plt.subplots(1,1,figsize=(6,4), sharey = False) \n", + "COL = sns.color_palette(n_colors = int(mat_trans_sort.shape[1]))\n", + "for i in range(mat_trans_sort.shape[1]):\n", + " ax.plot(QUANTS,np.sqrt(THETA_VARS[:,i]) , color = COL[i], label = KEYS[i], linewidth = 2)\n", + " ax.scatter(QUANTS,np.sqrt(THETA_VARS[:,i]) , color = COL[i])\n", + "ax.axhline(y=np.sqrt((1/12)), linestyle = '--', color = 'black', linewidth = 3)\n", + "ax.set_ylabel('Std. Dev.')\n", + "ax.title.set_text('Standard Deviation of Theta above Posterior Quantile')\n", + "ax.set_xlabel('Quantile')\n", + "ax.legend(loc = 'lower left', ncol = 3)\n", + "ax.set_ylim(0,np.sqrt((1/12))+0.05)\n", + "plt.show()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.8.18" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/examples/troubleshooting_convergence.ipynb b/examples/troubleshooting_convergence.ipynb index adfbf8e..c1c3a85 100644 --- a/examples/troubleshooting_convergence.ipynb +++ b/examples/troubleshooting_convergence.ipynb @@ -42,32 +42,37 @@ "\n", "3. Consider your modeling assumptions. A pooled analysis assumes that the same theta parameter value is shared across experiments, while a hierarchical analysis allows the theta values to differ between experiments. A clustered analysis assumes that there are discrete subpopulations of experiments with shared values. Does the model structure make sense for your data and for your analytical goals? Do you need to consider model discrepancy? Do the prior distribution hyperparameters make sense for your data? Below, we list some tunable hyperparameters that experienced users can modify:\n", "\n", - "* In the model call (e.g., ModelF, ModelMaterialStrength, etc.): \n", - " * s2: this option allows users to specify how and whether the experimental noise level should be estimated. MCMC sampler issues can sometimes be mitigated by fixing s2 to some user-specified value. If s2 = \"fix\", then the sd_est value specified in addVecExperiments is fixed for all MCMC iterations. \n", - "\n", - "* In addVecExperiments: \n", - " * sd_est: The initial values for the observation noise standard deviation\n", - " * s2_df: the initial values for the s2 Inverse Gamma prior degrees of freedom.\n", - " * discrep_tau: the fixed prior variance for the discrepancy basis coefficients controlling the amount of shrinkage toward zero\n", - "\n", - "* In setMCMC:\n", - " * decor: Number of iterations between each decorrelation step. Reducing this number may help hierarhical and clustered calibration to move around the parameter space at a steep computational cost\n", - " * start_var_theta: initial variance of adaptive MCMC proposal distributions for theta. Can be increased from default if posterior samples of theta are stuck at a single value across many iterations\n", - " * start_tau_theta: np.exp(start_tau_theta) is the initial scaling factor for the adaptive MCMC proposal covariance for theta. This can be tuned to modify acceptance rates, e.g., by making this smaller for samplers that get stuck at a single value across many iterations\n", - " * start_var_ls2: initial variance of adaptive MCMC proposal distributions for log(s2), i.e. the log of the observation error/noise standard deviation. Can be increased from default if posterior samples of theta are stuck at a single value across many iterations\n", - " * start_tau_ls2: np.exp(start_tau_ls2) is the initial scaling factor for the adaptive MCMC proposal covariance for log(s2). \n", - " * start_adapt_iter: MCMC iteration at which to start adapting the MCMC proposal distributions. \n", - "\n", - "* In setHierPriors: (Hierarchical model assumes experiment-specific thetai ~ Normal(theta0, Sigma0))\n", - " * theta0_prior_mean: the prior mean for the calibration parameter theta0. This can usually be left as default. \n", - " * theta0_prior_cov: the prior covariance for theta0, usually np.eye(self.p)*user_defined_prior_variance. This could be adjusted as needed. \n", - " * Sigma0_prior_df: prior degrees of freedom for the Inverse Wishart prior for Sigma0, at least 1 + self.p, where larger values generally indicate theta_i values closer to theta_0\n", - " * Sigma0_prior_scale: prior scale for the Inverse Wishart prior for Sigma0, where smaller values generally indicate theta_i values closer to theta_0\n", - "\n", - "* In setClusterPriors (Clustered model assumes experiment-speific thetai ~ DirichletProcess(G0, eta) where G0 = Normal(theta0, Sigma0)): \n", - " * nclustmax: the maximum number of unique theta values to estimate (i.e., maximum number of clusters). If this is too small, this can create sampling problems\n", - " * eta_prior_shape : prior shape for the Gamma prior for eta, the concentration parameter for the Dirichlet Process. Lower values indicate a higher propensity for experiments to cluster together. \n", - " * eta_prior_rate : prior rate for the Gamma prior for eta, the concentration parameter for the Dirichlet Process. Lower values indicate a higher propensity for experiments to cluster together. \n" + "* In the model call (e.g., **ModelF**, **ModelMaterialStrength**, etc.): \n", + "\n", + " * *s2*: this option allows users to specify how and whether the experimental noise level should be estimated. MCMC sampler issues can sometimes be mitigated by fixing s2 to some user-specified value. If *s2* = \"fix\", then the *sd_est* value specified in addVecExperiments is fixed for all MCMC iterations. \n", + "\n", + "* In **addVecExperiments**: \n", + "\n", + " * *sd_est*: The initial values for the observation noise standard deviation\n", + " * *s2_df*: the initial values for the s2 Inverse Gamma prior degrees of freedom.\n", + " * *discrep_tau*: the fixed prior variance for the discrepancy basis coefficients controlling the amount of shrinkage toward zero\n", + "\n", + "* In **setMCMC**:\n", + "\n", + " * *decor*: Number of iterations between each decorrelation step. Reducing this number may help hierarhical and clustered calibration to move around the parameter space at a steep computational cost\n", + " * *start_var_theta*: initial variance of adaptive MCMC proposal distributions for theta. Can be increased from default if posterior samples of theta are stuck at a single value across many iterations\n", + " * *start_tau_theta*: np.exp(*start_tau_theta*) is the initial scaling factor for the adaptive MCMC proposal covariance for theta. This can be tuned to modify acceptance rates, e.g., by making this smaller for samplers that get stuck at a single value across many iterations\n", + " * *start_var_ls2*: initial variance of adaptive MCMC proposal distributions for log(s2), i.e. the log of the observation error/noise standard deviation. Can be increased from default if posterior samples of theta are stuck at a single value across many iterations\n", + " * *start_tau_ls2*: np.exp(*start_tau_ls2*) is the initial scaling factor for the adaptive MCMC proposal covariance for log(s2). \n", + " * *start_adapt_iter*: MCMC iteration at which to start adapting the MCMC proposal distributions. \n", + "\n", + "* In **setHierPriors**: (Hierarchical model assumes experiment-specific thetai ~ Normal(theta0, Sigma0))\n", + "\n", + " * *theta0_prior_mean*: the prior mean for the calibration parameter theta0. This can usually be left as default. \n", + " * *theta0_prior_cov*: the prior covariance for theta0, usually np.eye(self.p)*user_defined_prior_variance. This could be adjusted as needed. \n", + " * *Sigma0_prior_df*: prior degrees of freedom for the Inverse Wishart prior for Sigma0, at least 1 + self.p, where larger values generally indicate theta_i values closer to theta_0\n", + " * *Sigma0_prior_scale*: prior scale for the Inverse Wishart prior for Sigma0, where smaller values generally indicate theta_i values closer to theta_0\n", + "\n", + "* In **setClusterPriors** (Clustered model assumes experiment-speific thetai ~ DirichletProcess(G0, eta) where G0 = Normal(theta0, Sigma0)): \n", + "\n", + " * *nclustmax*: the maximum number of unique theta values to estimate (i.e., maximum number of clusters). If this is too small, this can create sampling problems\n", + " * *eta_prior_shape* : prior shape for the Gamma prior for eta, the concentration parameter for the Dirichlet Process. Lower values indicate a higher propensity for experiments to cluster together. \n", + " * *eta_prior_rate* : prior rate for the Gamma prior for eta, the concentration parameter for the Dirichlet Process. Lower values indicate a higher propensity for experiments to cluster together. \n" ] }, { @@ -119,13 +124,13 @@ "source": [ "### My calibration is taking a really long time!\n", "\n", - "Calibrations with my addVecExperiments calls to set up the calibration object can often run much more slowly. This is because this casts the observed data into a structure that does not leverage vectorized model evaluations. If possible, we recommend recasting multiple data vectors into a single long data vector entered into the model using a single addVecExperiments call. The IMPALA model object used to obtain computer model predictions would also have to be modified to provide predictions for the longer observation vector (e.g., by using ModelF to write a custom computer model prediction function). \n", + "Calibrations with my **addVecExperiments** calls to set up the calibration object can often run much more slowly. This is because this casts the observed data into a structure that does not leverage vectorized model evaluations. If possible, we recommend recasting multiple data vectors into a single long data vector entered into the model using a single **addVecExperiments** call. The IMPALA model object used to obtain computer model predictions would also have to be modified to provide predictions for the longer observation vector (e.g., by using **ModelF** to write a custom computer model prediction function). \n", "\n", - "If you are calibrating using a custom computer model evaluated using ModelF, consider whether your computer model is too slow and whether an emulator is needed to speed up the computation.\n", + "If you are calibrating using a custom computer model evaluated using **ModelF**, consider whether your computer model is too slow and whether an emulator is needed to speed up the computation.\n", "\n", - "Calibrations using calibClust tend to be substantially slower than their hierarchical and pooled counterparts. This is a topic of continued code development. \n", + "Calibrations using **calibClust** tend to be substantially slower than their hierarchical and pooled counterparts. This is a topic of continued code development. \n", "\n", - "For very large observed datasets, the current specification of ModelF can be slow. See ModelF_bigdata for a version tailored to the bigger data setting. " + "For very large observed datasets, the current specification of **ModelF** can be slow. See **ModelF_bigdata** for a version tailored to the bigger data setting. " ] } ],