robinthibaut / deep-cca

Deep Canonical Correlation Analysis with Python

DCCA: Deep Canonical Correlation Analysis

I'm currently adapting the code from https://github.com/VahidooX/DeepCCA.

Objectives

• Solving compatibility-issues with anterior version of Keras
• Make it compatible with Tensorflow backend (remove Theano dependency)
• Make it compatible with scikit-learn framework
• Currently, the code is designed to solve a classification problem - I'm interested in solving regression problems
• Figure out how to back-transform the data that went through the NN layers to their original space.

Description

This is an implementation of Deep Canonical Correlation Analysis in Python.

DCCA is a non-linear version of CCA which uses neural networks as the mapping functions instead of linear transformers. DCCA is originally proposed in the following paper:

Galen Andrew, Raman Arora, Jeff Bilmes, Karen Livescu, "Deep Canonical Correlation Analysis.", ICML, 2013.

It uses the Keras library with the Tensorflow backend.

Differences with the original paper

The following are the differences between this implementation and the original paper:

• The non-saturating version of sigmoid is substituted by another non-saturating activation function (ReLU).
• Pre-training is not done in this implementation.