Simple Exponential Family PCA

Jun Li, Dacheng Tao
; Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, JMLR Workshop and Conference Proceedings 9:453-460, 2010.

Abstract

Bayesian principal component analysis (BPCA), a probabilistic reformulation of PCA with Bayesian model selection, is a systematic approach to determining the number of essential principal components (PCs) for data representation. However, it assumes that data are Gaussian distributed and thus it cannot handle all types of practical observations, e.g. integers and binary values. In this paper, we propose simple exponential family PCA (SePCA), a generalised family of probabilistic principal component analysers. SePCA employs exponential family distributions to handle general types of observations. By using Bayesian inference, SePCA also automatically discovers the number of essential PCs. We discuss techniques for fitting the model, develop the corresponding mixture model, and show the effectiveness of the model based on experiments.

Cite this Paper


BibTeX
@InProceedings{pmlr-v9-li10b, title = {Simple Exponential Family PCA}, author = {Jun Li and Dacheng Tao}, booktitle = {Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics}, pages = {453--460}, year = {2010}, editor = {Yee Whye Teh and Mike Titterington}, volume = {9}, series = {Proceedings of Machine Learning Research}, address = {Chia Laguna Resort, Sardinia, Italy}, month = {13--15 May}, publisher = {JMLR Workshop and Conference Proceedings}, pdf = {http://proceedings.mlr.press/v9/li10b/li10b.pdf}, url = {http://proceedings.mlr.press/v9/li10b.html}, abstract = {Bayesian principal component analysis (BPCA), a probabilistic reformulation of PCA with Bayesian model selection, is a systematic approach to determining the number of essential principal components (PCs) for data representation. However, it assumes that data are Gaussian distributed and thus it cannot handle all types of practical observations, e.g. integers and binary values. In this paper, we propose simple exponential family PCA (SePCA), a generalised family of probabilistic principal component analysers. SePCA employs exponential family distributions to handle general types of observations. By using Bayesian inference, SePCA also automatically discovers the number of essential PCs. We discuss techniques for fitting the model, develop the corresponding mixture model, and show the effectiveness of the model based on experiments.} }
Endnote
%0 Conference Paper %T Simple Exponential Family PCA %A Jun Li %A Dacheng Tao %B Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2010 %E Yee Whye Teh %E Mike Titterington %F pmlr-v9-li10b %I PMLR %J Proceedings of Machine Learning Research %P 453--460 %U http://proceedings.mlr.press %V 9 %W PMLR %X Bayesian principal component analysis (BPCA), a probabilistic reformulation of PCA with Bayesian model selection, is a systematic approach to determining the number of essential principal components (PCs) for data representation. However, it assumes that data are Gaussian distributed and thus it cannot handle all types of practical observations, e.g. integers and binary values. In this paper, we propose simple exponential family PCA (SePCA), a generalised family of probabilistic principal component analysers. SePCA employs exponential family distributions to handle general types of observations. By using Bayesian inference, SePCA also automatically discovers the number of essential PCs. We discuss techniques for fitting the model, develop the corresponding mixture model, and show the effectiveness of the model based on experiments.
RIS
TY - CPAPER TI - Simple Exponential Family PCA AU - Jun Li AU - Dacheng Tao BT - Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics PY - 2010/03/31 DA - 2010/03/31 ED - Yee Whye Teh ED - Mike Titterington ID - pmlr-v9-li10b PB - PMLR SP - 453 DP - PMLR EP - 460 L1 - http://proceedings.mlr.press/v9/li10b/li10b.pdf UR - http://proceedings.mlr.press/v9/li10b.html AB - Bayesian principal component analysis (BPCA), a probabilistic reformulation of PCA with Bayesian model selection, is a systematic approach to determining the number of essential principal components (PCs) for data representation. However, it assumes that data are Gaussian distributed and thus it cannot handle all types of practical observations, e.g. integers and binary values. In this paper, we propose simple exponential family PCA (SePCA), a generalised family of probabilistic principal component analysers. SePCA employs exponential family distributions to handle general types of observations. By using Bayesian inference, SePCA also automatically discovers the number of essential PCs. We discuss techniques for fitting the model, develop the corresponding mixture model, and show the effectiveness of the model based on experiments. ER -
APA
Li, J. & Tao, D.. (2010). Simple Exponential Family PCA. Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, in PMLR 9:453-460

Related Material