Non-Gaussian Discriminative Factor Models via the Max-Margin Rank-Likelihood

Xin Yuan, Ricardo Henao, Ephraim Tsalik, Raymond Langley, Lawrence Carin
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:1254-1263, 2015.

Abstract

We consider the problem of discriminative factor analysis for data that are in general non-Gaussian. A Bayesian model based on the ranks of the data is proposed. We first introduce a max-margin version of the rank-likelihood. A discriminative factor model is then developed, integrating the new max-margin rank-likelihood and (linear) Bayesian support vector machines, which are also built on the max-margin principle. The discriminative factor model is further extended to the nonlinear case through mixtures of local linear classifiers, via Dirichlet processes. Fully local conjugacy of the model yields efficient inference with both Markov Chain Monte Carlo and variational Bayes approaches. Extensive experiments on benchmark and real data demonstrate superior performance of the proposed model and its potential for applications in computational biology.

Cite this Paper


BibTeX
@InProceedings{pmlr-v37-yuan15, title = {Non-Gaussian Discriminative Factor Models via the Max-Margin Rank-Likelihood}, author = {Yuan, Xin and Henao, Ricardo and Tsalik, Ephraim and Langley, Raymond and Carin, Lawrence}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {1254--1263}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/yuan15.pdf}, url = {https://proceedings.mlr.press/v37/yuan15.html}, abstract = {We consider the problem of discriminative factor analysis for data that are in general non-Gaussian. A Bayesian model based on the ranks of the data is proposed. We first introduce a max-margin version of the rank-likelihood. A discriminative factor model is then developed, integrating the new max-margin rank-likelihood and (linear) Bayesian support vector machines, which are also built on the max-margin principle. The discriminative factor model is further extended to the nonlinear case through mixtures of local linear classifiers, via Dirichlet processes. Fully local conjugacy of the model yields efficient inference with both Markov Chain Monte Carlo and variational Bayes approaches. Extensive experiments on benchmark and real data demonstrate superior performance of the proposed model and its potential for applications in computational biology.} }
Endnote
%0 Conference Paper %T Non-Gaussian Discriminative Factor Models via the Max-Margin Rank-Likelihood %A Xin Yuan %A Ricardo Henao %A Ephraim Tsalik %A Raymond Langley %A Lawrence Carin %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-yuan15 %I PMLR %P 1254--1263 %U https://proceedings.mlr.press/v37/yuan15.html %V 37 %X We consider the problem of discriminative factor analysis for data that are in general non-Gaussian. A Bayesian model based on the ranks of the data is proposed. We first introduce a max-margin version of the rank-likelihood. A discriminative factor model is then developed, integrating the new max-margin rank-likelihood and (linear) Bayesian support vector machines, which are also built on the max-margin principle. The discriminative factor model is further extended to the nonlinear case through mixtures of local linear classifiers, via Dirichlet processes. Fully local conjugacy of the model yields efficient inference with both Markov Chain Monte Carlo and variational Bayes approaches. Extensive experiments on benchmark and real data demonstrate superior performance of the proposed model and its potential for applications in computational biology.
RIS
TY - CPAPER TI - Non-Gaussian Discriminative Factor Models via the Max-Margin Rank-Likelihood AU - Xin Yuan AU - Ricardo Henao AU - Ephraim Tsalik AU - Raymond Langley AU - Lawrence Carin BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-yuan15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 1254 EP - 1263 L1 - http://proceedings.mlr.press/v37/yuan15.pdf UR - https://proceedings.mlr.press/v37/yuan15.html AB - We consider the problem of discriminative factor analysis for data that are in general non-Gaussian. A Bayesian model based on the ranks of the data is proposed. We first introduce a max-margin version of the rank-likelihood. A discriminative factor model is then developed, integrating the new max-margin rank-likelihood and (linear) Bayesian support vector machines, which are also built on the max-margin principle. The discriminative factor model is further extended to the nonlinear case through mixtures of local linear classifiers, via Dirichlet processes. Fully local conjugacy of the model yields efficient inference with both Markov Chain Monte Carlo and variational Bayes approaches. Extensive experiments on benchmark and real data demonstrate superior performance of the proposed model and its potential for applications in computational biology. ER -
APA
Yuan, X., Henao, R., Tsalik, E., Langley, R. & Carin, L.. (2015). Non-Gaussian Discriminative Factor Models via the Max-Margin Rank-Likelihood. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:1254-1263 Available from https://proceedings.mlr.press/v37/yuan15.html.

Related Material