Deep Canonically Correlated LSTMs
This work is incremental, building on existing Deep Canonical Correlation methods by applying them to sequence data with LSTMs.
The authors tackled the problem of learning nonlinear transformations of variable-length sequences to embed them into a correlated, fixed-dimensional space, using Deep Canonically Correlated LSTMs to extend previous work on Deep Canonical Correlation by incorporating temporal relationships in multi-view time-series data.
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.