ABSTRACT
Clustering data in high dimensions is believed to be a hard problem in general. A number of efficient clustering algorithms developed in recent years address this problem by projecting the data into a lower-dimensional subspace, e.g. via Principal Components Analysis (PCA) or random projections, before clustering. Here, we consider constructing such projections using multiple views of the data, via Canonical Correlation Analysis (CCA).
Under the assumption that the views are un-correlated given the cluster label, we show that the separation conditions required for the algorithm to be successful are significantly weaker than prior results in the literature. We provide results for mixtures of Gaussians and mixtures of log concave distributions. We also provide empirical support from audio-visual speaker clustering (where we desire the clusters to correspond to speaker ID) and from hierarchical Wikipedia document clustering (where one view is the words in the document and the other is the link structure).
- Achlioptas, D., & McSherry, F. (2005). On spectral learning of mixtures of distributions. Conf. on Learning Thy (pp. 458--469). Google ScholarDigital Library
- Ando, R. K., & Zhang, T. (2007). Two-view feature generation model for semi-supervised learning. Int. Conf. on Machine Learning (pp. 25--32). Google ScholarDigital Library
- Arora, S., & Kannan, R. (2005). Learning mixtures of separated nonspherical Gaussians. Ann. Applied Prob., 15, 69--92.Google ScholarCross Ref
- Blaschko, M. B., & Lampert, C. H. (2008). Correlational spectral clustering. Conf. on Comp. Vision and Pattern Recognition.Google ScholarCross Ref
- Blum, A., & Mitchell, T. (1998). Combining labeled and unlabeled data with co-training. Conf. on Learning Thy. (pp. 92--100). Google ScholarDigital Library
- Brubaker, S. C., & Vempala, S. (2008). Isotropic PCA and affine-invariant clustering. Found. of Comp. Sci. (pp. 551--560). Google ScholarDigital Library
- Chaudhuri, K., & Rao, S. (2008). Learning mixtures of distributions using correlations and independence. Conf. On Learning Thy. (pp. 9--20).Google Scholar
- Dasgupta, S. (1999). Learning mixtures of Gaussians. Found. of Comp. Sci. (pp. 634--644). Google ScholarDigital Library
- Dasgupta, S., & Schulman, L. (2000). A two-round variant of EM for Gaussian mixtures. Uncertainty in Art. Int. (pp. 152--159). Google ScholarDigital Library
- Davis, S. B., & Mermelstein, P. (1980). Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences. IEEE Trans. Acoustics, Speech, and Signal Proc., 28, 357--366.Google ScholarCross Ref
- Dunn, G., & Everitt, B. (2004). An introduction to math. taxonomy. Dover Books.Google Scholar
- Kakade, S. M., & Foster, D. P. (2007). Multi-view regression via canonical correlation analysis. Conf. Learning Thy (pp. 82--96). Google ScholarDigital Library
- Kannan, R., Salmasian, H., & Vempala, S. (2005). The spectral method for general mixture models. Conf. on Learning Thy (pp. 444--457). Google ScholarDigital Library
- Rudelson, M., & Vershynin, R. (2007). Sampling from large matrices: An approach through geometric functional analysis. Jour. of ACM. Google ScholarDigital Library
- Sanderson, C. (2008). Biometric person recognition: Face, speech and fusion. VDM-Verlag.Google Scholar
- Vempala, V., & Wang, G. (2002). A spectral algorithm for learning mixtures of distributions. Found. of Comp. Sci. (pp. 113--123). Google ScholarDigital Library
Index Terms
- Multi-view clustering via canonical correlation analysis
Recommendations
Multi-view dimensionality reduction via canonical random correlation analysis
Canonical correlation analysis (CCA) is one of the most well-known methods to extract features from multi-view data and has attracted much attention in recent years. However, classical CCA is unsupervised and does not take discriminant information into ...
A New Locality-Preserving Canonical Correlation Analysis Algorithm for Multi-View Dimensionality Reduction
Canonical correlation analysis (CCA) is a well-known technique for extracting linearly correlated features from multiple views (i.e., sets of features) of data. Recently, a locality-preserving CCA, named LPCCA, has been developed to incorporate the ...
Multi-view Fractional Deep Canonical Correlation Analysis for Subspace Clustering
Neural Information ProcessingAbstractCanonical correlation analysis (CCA) is a classic unsupervised dimensionality reduction approach, but it has difficulty to analyze the nonlinear relationship and learn from more than two views. In addition, real-world data sets often contain much ...
Comments