Deep Canonical Correlation Analysis

[jinglescode]

Paper

Link: http://proceedings.mlr.press/v28/andrew13.pdf
Year: 2013
Video: https://www.youtube.com/watch?v=mcS8dysEWPw
Code: https://github.com/Michaelvll/DeepCCA

Summary

• learn complex nonlinear transformations of two views of data such that the resulting representations are highly linearly correlated
• significantly higher correlation than those learned by CCA and KCCA
• introduce a novel non-saturating sigmoid function based on the cube root

Methods

• Deep CCA computes representations of the two views by passing them through multiple stacked layers of nonlinear transformation
• The goal is to jointly learn parameters for both views such that the correlation is as high as possible

Results

• obtain improved representations with respect to the correlation objective measured on unseen data