infancy

Building an MLP Neural Network from scratch, using only numpy. My main goal is to reduce the degree of "black-box" understanding with NNs, specifically back-propagation.

RUNNING & TRAINING THE NETWORK

To run:

1. Clone the repo
2. Run python testingEnvironment.py

PERFORMANCE

Epochs (iterations): 400
Average loss after training: 14.65~
Accuracy: // TODO: ADD ACCURACY TESTING

DESIGN

You can read my detailed thoughts here: Notion Page

The following is a visualization of the forward propogation alg for each layer, but note that the activation functions might change