This is a Neural Network built from scratch using only Numpy for numerical computation.

In order to train the model you will need to install some dependencies:

pip install -r requirements.txt

Once you're done with that, you can start training the model:

python main.py --epochs 20 --lr 0.001 --batch-size 64

As this is just an example implementation we'll be using the standard benchmark MNIST dataset of handwritten digits for classification to train our model. You can either load the data using Keras or download it from here. However I'll be using the Keras Dataset helper.