July 26, 2016

An artificial neural network using Keras

  • supervised learning
  • artificial neural networks
  • keras

Python has no shortage of utilities for building artificial neural networks (ANNs). Keras is a library for building ANNs that sits atop of Theano or TensorFlow. I recently tried it out when building a submission for Kaggle’s Ultrasound Nerve Segmentation competition and found it to be a pleasure to work with. I find it strikes a nice balance — allowing you to focus on the network architecture and avoid getting bogged down in the implementation details. However you can still get your hands dirty if you need to.

I’ve put together a notebook that uses Keras to implement the same ANN I coded from scratch in a previous notebook. As usual, if you want to follow along or play with the code yourself you can find the notebook in my ml-python repo.

Artificial neural network in Python with Keras