Skip to content

Latest commit

 

History

History
 
 

tensorflow-mnist-tutorial

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Image

This is support code for the codelab "Tensorflow and deep learning - without a PhD"

The presentation explaining the underlying concepts is here.

The lab takes 2.5 hours and takes you through the design and optimisation of a neural network for recognising handwritten digits, from the simplest possible solution all the way to a recognition accuracy above 99%. It covers dense and convolutional networks, as well as techniques such as learning rate decay and dropout.

Installation instructions here. The short version is: install Python3, then pip3 install tensorflow and matplotlib.

The most advanced advanced neural network in this repo achieves 99.5% accuracy on the MNIST dataset (world best is 99.7%) and uses batch normalization.

This lab uses low-level Tensorflow because it is intended as a starting point for developers learning neural network techniques. It is important to see what is going on with trainable variables (weights and biases) before moving to higher-level APIs that hide these concepts. If you are looking for a high-level Tensorflow sample using layers, Estimator and Dataset APIs, you will find it in the mlengine folder.


Disclaimer: This is not an official Google product but sample code provided for an educational purpose