Sigmoid toolkit

This is a repo with the numerical scripts used in the study of the SM-dg family of curves.

Installation and dependencies

pip

These scripts depend on a few numerical python libraries like scipy and numpy. To install them you need the python module pip.

To obtain pip, please follow the guide at:

https://pip.pypa.io/en/stable/installation/

Requirements

After obtaining a functional instance of pip, in the trunk of repo you will find a file named requirements.txt.

To install all the needed libraries to run these scripts open a terminal and run:

pip3 install -r requirements.txt

Wait for the installation of all libraries to finish.

Configuration

To configure the general behaiviour of all scripts that currently exist and might be developed in the future, please use the config.toml file available in the repo. It's general structure should look something like this:

t0 = 0
t_final = 10000
initial_alpha = 1e-8
n_terms_taylor = 1
csv_delimiter = "\t"

[param_finder_props]
d_min = 1.0
g_min = 0.5
mu_min = 0

d_max = 3.0000000002
g_max = 2.50000002
mu_max = 5

g_ini = 1.000000001
d_ini = 2.0000000001
mu_ini = 1.3301409949454641e-06

shift_time = 0

# 0 to skip plotting, > 0 to plot
plot_graphs_matplotlib = 1

[param_finder_props.filters]
# Data filters
alpha_min = 0
alpha_max = 1

[time_offset_finder_proprs]
fit_procedure = "nlsq"

t0 и t_final set the integration interval for the numerical integrator.

Scripts

Basic usage (with autorun.py)

There is now a script autorun.py that automates most of the configuration, except for the bounds and initial conditions

1. Go to the config.toml and set bounds for d,g, mu (= 1/ tau) and save the file.

2. Run python3 autorun.py <path_to_dataset.tsv>

3. Done.

Parameter finder

This script uses non-linear least squares to fit the best values for D and g for tab-separated dataset.

Usage

python3 parameter_finder.py <path_to_input_file.csv>

The first column in the tab-seperated list should be the values of the time coordinate and the second - the values for alpha.