A kernel method for canonical correlation analysis
read the original abstract
Canonical correlation analysis is a technique to extract common features from a pair of multivariate data. In complex situations, however, it does not extract useful features because of its linearity. On the other hand, kernel method used in support vector machine is an efficient approach to improve such a linear method. In this paper, we investigate the effectiveness of applying kernel method to canonical correlation analysis.
This paper has not been read by Pith yet.
Forward citations
Cited by 3 Pith papers
-
Sliding Window Informative Canonical Correlation Analysis
SWICCA extends canonical correlation analysis to streaming data by pairing a streaming PCA backend with a sliding window of samples, supported by simulations and a theoretical performance guarantee.
-
Complex Interpolation of Matrices with an application to Multi-Manifold Learning
Exact log-linearity of the operator norm of the matrix interpolation A^{1-x} B^x is generically equivalent to the matrices sharing an eigenvector, providing theoretical support for a multi-manifold learning method on ...
-
Feature Extraction in the Remote Sensing Data Value Chain: A Systematic Review of Methods and Applications
A systematic review that introduces a framework for feature extraction in remote sensing, traces its evolution in the data value chain, and synthesizes trends toward unified representations and foundation models.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.