Large scale canonical correlation analysis with iterative least squares

TL;DR

This paper introduces L-CCA, an iterative algorithm that efficiently computes Canonical Correlation Analysis on large sparse datasets, with proven convergence and superior performance over existing methods.

Contribution

The paper presents L-CCA, a novel iterative method for fast large-scale CCA computation with theoretical convergence guarantees and improved accuracy.

Findings

L-CCA converges asymptotically and in finite time.

L-CCA outperforms existing approximation schemes on real datasets.

L-CCA is effective for large sparse datasets.

Abstract

Canonical Correlation Analysis (CCA) is a widely used statistical tool with both well established theory and favorable performance for a wide range of machine learning problems. However, computing CCA for huge datasets can be very slow since it involves implementing QR decomposition or singular value decomposition of huge matrices. In this paper we introduce L-CCA, a iterative algorithm which can compute CCA fast on huge sparse datasets. Theory on both the asymptotic convergence and finite time accuracy of L-CCA are established. The experiments also show that L-CCA outperform other fast CCA approximation schemes on two real datasets.