PyNN

A pure numpy implementation of layers, learning algorithms and objective functions for neural networks.

Layers and Activation Functions

Layers:

1. Linear
2. Convolution (in-progress)

Activation Functions:

1. Sigmoid
2. Softmax
3. ReLU
4. Tanh (in-progress)
5. LeakyReLU

Learning Algorithms

1. Stochastic Mini-Batch Gradient Descent

Metrics

1. Area Under the Curve (AUC)

Objective Functions

1. Mean Squared Error
2. Categorical Cross Entropy