%0 Conference Paper %1 pmlr-v28-chatzis13 %A Chatzis, Sotirios %B Proceedings of the 30th International Conference on Machine Learning %C Atlanta, Georgia, USA %D 2013 %E Dasgupta, Sanjoy %E McAllester, David %I PMLR %K bayesian hierarchical max-margin non-parametric readings variational %N 3 %P 729--737 %T Infinite Markov-Switching Maximum Entropy Discrimination Machines %U http://proceedings.mlr.press/v28/chatzis13.html %V 28 %X In this paper, we present a method that combines the merits of Bayesian nonparametrics, specifically stick-breaking priors, and large-margin kernel machines in the context of sequential data classification. The proposed model postulates a set of (theoretically) infinite interdependent large-margin classifiers as model components, that robustly capture local nonlinearity of complex data. The postulated large-margin classifiers are connected in the context of a Markov-switching construction that allows for capturing complex temporal dynamics in the modeled datasets. Appropriate stick-breaking priors are imposed over the component switching mechanism of our model to allow for data-driven determination of the optimal number of component large-margin classifiers, under a standard nonparametric Bayesian inference scheme. Efficient model training is performed under the maximum entropy discrimination (MED) framework, which integrates the large-margin principle with Bayesian posterior inference. We evaluate our method using several real-world datasets, and compare it to state-of-the-art alternatives.

@inproceedings{pmlr-v28-chatzis13, abstract = {In this paper, we present a method that combines the merits of Bayesian nonparametrics, specifically stick-breaking priors, and large-margin kernel machines in the context of sequential data classification. The proposed model postulates a set of (theoretically) infinite interdependent large-margin classifiers as model components, that robustly capture local nonlinearity of complex data. The postulated large-margin classifiers are connected in the context of a Markov-switching construction that allows for capturing complex temporal dynamics in the modeled datasets. Appropriate stick-breaking priors are imposed over the component switching mechanism of our model to allow for data-driven determination of the optimal number of component large-margin classifiers, under a standard nonparametric Bayesian inference scheme. Efficient model training is performed under the maximum entropy discrimination (MED) framework, which integrates the large-margin principle with Bayesian posterior inference. We evaluate our method using several real-world datasets, and compare it to state-of-the-art alternatives. }, added-at = {2020-02-17T03:24:13.000+0100}, address = {Atlanta, Georgia, USA}, author = {Chatzis, Sotirios}, biburl = {https://www.bibsonomy.org/bibtex/203485cda7bd3f6c95a5bcc99ddb4900a/kirk86}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, description = {Infinite Markov-Switching Maximum Entropy Discrimination Machines}, editor = {Dasgupta, Sanjoy and McAllester, David}, interhash = {a1b43242cfa77af01f80e548690a004e}, intrahash = {03485cda7bd3f6c95a5bcc99ddb4900a}, keywords = {bayesian hierarchical max-margin non-parametric readings variational}, month = {17--19 Jun}, number = 3, pages = {729--737}, pdf = {http://proceedings.mlr.press/v28/chatzis13.pdf}, publisher = {PMLR}, series = {Proceedings of Machine Learning Research}, timestamp = {2020-02-17T03:24:13.000+0100}, title = {Infinite Markov-Switching Maximum Entropy Discrimination Machines}, url = {http://proceedings.mlr.press/v28/chatzis13.html}, volume = 28, year = 2013 }