Deep Canonically Correlated LSTMs

TL;DR

This paper introduces Deep Canonically Correlated LSTMs, a method that learns nonlinear transformations of multi-view time-series data to embed them into a correlated, fixed-dimensional space, capturing temporal relationships.

Contribution

It extends Deep Canonical Correlation Analysis by incorporating LSTMs to handle variable-length sequences and temporal dependencies in multi-view data.

Findings

Effective nonlinear embedding of time-series data.

Captures temporal relationships within multi-view sequences.

Improves correlation-based representations for sequential data.

Abstract

We examine Deep Canonically Correlated LSTMs as a way to learn nonlinear transformations of variable length sequences and embed them into a correlated, fixed dimensional space. We use LSTMs to transform multi-view time-series data non-linearly while learning temporal relationships within the data. We then perform correlation analysis on the outputs of these neural networks to find a correlated subspace through which we get our final representation via projection. This work follows from previous work done on Deep Canonical Correlation (DCCA), in which deep feed-forward neural networks were used to learn nonlinear transformations of data while maximizing correlation.