This post is a Google Colab notebook showing how to visualize a neural network as it trains. It’s personally been really useful to me as a convenient tool for both debugging and building intuition. Shown below is a simple feedforward neural network attempting to fit the function f(x,y) = x^2 - cos(y)

https://deepmind.com/research/publications/Artificial-Intelligence-Values-and-Alignment

What’s all of this about?