Neural Network (Self-)Tutorial

Back in 2017, I took Andrew Ng’s Coursera class in machine learning. I was inspired to better establish my knowledge of what he covered.

The code contained represents my success at creating and estimating fully-connected neural network models (however inefficient). This firmed up my knowledge of:

• python
• numpy
• neural networks–the fully connected (a.k.a. dense) variety
• backpropigation

I created the following modules:

• inputLayers
• hiddenLayers
• outputLayers
• activation_functions

These are called in the module

• neuralNets

All of the above are packaged into nnfiles, which is called in MNIST test.ipynb to estimate a fully-connected neural network on the MNIST dataset. I was able to achieve 95% accuracy!

This was among my first projects in python and git.

Several years later, I added the code necessary to make each file module and provided a virtual environment (along with a handful of necessary updates).