This repository is intended to include custom utilities designed to facilitate effective artificial neural networks design.
These utility functions and network layers could then be used across various neural network applications.
The codebase is in Python, leveraging Keras backend and tensorflow.
Current contents include:
- factorization machine layer allowing parametrized support for specifying 2-way interactions
- wrapper around Mixture of Experts model designed by TensorFlow team.
- slicing layer to facilitate providing data to keras as a data frame or numpy array
Examples to follow shortly. Please reach out to me if you have suggestions or observations.
Enjoy, Pasha