This repo is derived from the blog by @iamtrask.
I added and changed a few things for educational purposes. In particular, I spelled out key functions, modified the update rule, and plotted out the losses and weights. I also added "rev" scripts which flip the matrices and dot products.
As @iamtrask mentioned, I found it very helpful to try writing these networks from memory.
Each script has an argument for plotting:
python two_layer_nn.py --plot
python three_layer_nn.py --plot
