baubels / numpynets

Neural nets from scratch using NumPy.

Neural nets from scratch using NumPy.

Usage

1. clone repo
2. install locally (python -m pip install -e ./numpynets)
3. get 2d image data

x_train.shape = (n, height, width), y_train.shape = (n, classes) or in addition provide x_valid.shape=(m, height, width), y_valid.shape(m, classes)

4. initialise and run an arbitrary length feed-forward fully connected net

With the learnt net:

1. extract learnt weights/biases with .trained_ned[layer_num].W, .trained_net[layer_num].B
2. extract learning histories/losses using .history
3. predict values for new inputs with .predict(xdata)

Network Specs

• He initialisations
• (Stochastic) Gradient Descent
• Feed-forward and fully connected

To implement

• convs
• 1d data as input (req's a minor bug fix)
• easier custom loss, activations, initialisations
• cuda-aware training (this has been done just needs uploading)