Publications

2011

---

Crowdclustering: Using crowdsourcing to discover categories in collections of images (and other human-interpretable patterns.)

• R. Gomes, P. Welinder, A. Krause, and P. Perona (2011). Crowdclustering. Advances in Neural Information Processing Systems.
  Paper Bibtex Extended Technical Report Code

Ph.D. Thesis: Can we build automatic categorization systems that learn and add categories over time with minimal human supervision?

• R. Gomes (2011). Towards Open Ended Learning: Budgets, Model Selection, and Representation.
  Thesis Bibtex

2010

---

Clustering via unsupervised learning of probabilistic discriminative classifiers (kernelized logistic regression).

• R. Gomes, A. Krause, and P. Perona (2010). Disciminative Clustering by Regularized Information Maximization. Advances in Neural Information Processing Systems.
  Paper Bibtex Code

Near optimal selection of informative examples from data streams. These data examples may be used in nonparametric clustering or regression problems.

• R. Gomes and A. Krause (2010). Budgeted Nonparametric Learning from Data Streams. Proceedings of the International Conference on Machine Learning.
  Paper Bibtex Long Version

2008

---

Incremental/online algorithm for mixture model clustering with the Dirichlet process mixture model. Automatically adjusts the number of clusters as evidence arrives. The method is based on variational approximate inference.

• R. Gomes, M. Welling, and P. Perona (2008). Incremental learning of nonparametric Bayesian mixture models. Proceedings of Computer Vision and Pattern Recognition.
  Paper Extended Thesis Chapter Bibtex Code

Extends the techniques of the above paper to the topic model, an hierarchical extension of the mixture model suited to modeling documents and images.

• R. Gomes, M. Welling, and P. Perona (2008). Memory bounded inference in topic models. Proceedings of the International Conference of Machine Learning.
  Paper Bibtex Video Lecture Slides
  Note: Typos were corrected in this version of the paper. In a number of places the sufficient statistics \phi_{l}(x) were written instead as \phi_{kl}(x). (There is no dependence on k).