Real Time Activation Visualization of Hidden Layers

Instructions:

- Use finger to draw a digit within the red box

- Click on a neuron in the first hidden layer so see weights connecting it to the input layer

- Try changing the regularization method or cost function to see effect on the weights of the network

Large
Positive Weight Large Negative
Weight

This visualization is inspired by a demo I encounterd while working through the Briliant.org Neural Network Course

The neural network was built from scratch through the direction of the book Neural Networks and Deep Learning by Michael Nielsen

This visualization demonstrates the the feedforward activation of a network while classifying handwritten digits.

The network uses the real weights and biases generated from training the network on the MNIST dataset.

To see the how the network was implemented, including stochastic gradient descent and backpropogation, see the jupyter notebook here

Resources that helped me a lot with understanding the backpropogation algorithm:
ML Glossary
A Step by Step Backpropagation Example
Diagram for visualizing Backprop