Summary

My thesis work studies methods for doing exact and efficient machine learning from network data while explicitly modeling the structural measure of degree.

More generally, I spend my days thinking about machine learning for relational data, where traditional methods are often insufficient. Among other things, I am exploring approaches to properly bridge relational data and the powerful mathematical formalisms of graphs and combinatorics that we use to algorithmically learn from it.

Refereed Journal Papers

Machine Learning for the New York City Power Grid.
Cynthia Rudin, David Waltz, Roger Anderson, Albert Boulanger, Ansaf Salleb-Aouissi, Maggie Chow, Haimonti Dutta, Philip Gross, Bert Huang, Steve Ierome, Delfina Isaac, Artie Kressner, Rebecca Passonneau, Axinia Radeva, and Leon Wu. IEEE Transactions on Pattern Analysis and Machine Intelligence. To appear.

Refereed Conference Papers

Fast b-Matching via Sufficient Selection Belief Propagation
Bert Huang and Tony Jebara. Fourteenth International Conference on Artificial Intelligence and Statistics (AISTATS) 2011.
PDF, code, poster.

Collaborative Filtering via Rating Concentration
Bert Huang and Tony Jebara. Thirteenth International Conference on Artificial Intelligence and Statistics (AISTATS) 2010.
PDF, code directory, code archive, poster

Exact Graph Structure Estimation with Degree Priors
Bert Huang and Tony Jebara. International Conference on Machine Learning and Applications (ICMLA) 2009.
PDF

Alive on Back-feed Culprit Identifaction via Machine Learning
Bert Huang, Ansaf Salleb-Aouissi, and Phil Gross. International Conference on Machine Learning and Applications (ICMLA) 2009. Special Session on Machine Learning in Energy Applications.
PDF

Discovering Characterization Rules from Rankings
Ansaf Salleb-Aouissi, Bert Huang, and David Waltz. International Conference on Machine Learning and Applications (ICMLA) 2009.
PDF

Maximum Entropy Density Estimation with Incomplete Presence-Only Data.
Bert Huang and Ansaf Salleb-Aouissi. AISTATS 2009.
PDF, poster

Vers des Machines a Vecteurs de Support "Actionables": Une Approche Fondee sur le Classement. (Toward Actionable Support Vector Machines: A Ranking Based Approach)
Ansaf Salleb-Aouissi, Bert Huang, and David Waltz. Knowledge Extraction and Management (Extraction et Gestion des Connaissances) EGC 2008, Sophia Antipolis, France. Best Paper Award.

Loopy Belief Propagation for Bipartite Maximum Weight b-Matching.
Bert Huang and Tony Jebara. AISTATS 2007.
PDF, PS, slides, code.

Refereed Workshop Papers and Abstracts

Network Prediction with Degree Distributional Metric Learning
Bert Huang, Blake Shaw, and Tony Jebara. Interdisciplinary Workshop on Information and Decision in Social Networks (WIDS 2011).
Poster and abstract

Learning with Subgraph Estimation and Degree Priors.
Bert Huang and Tony Jebara. New York Academy of Sciences Machine Learning Symposium, November 2009.
PDF

Maximum Likelihood Graph Structure Estimation with Degree Distributions.
Bert Huang and Tony Jebara. Analyzing Graphs: Theory and Applications, NIPS Workshop, December 2008.
PDF

Approximating the Permanent with Belief Propagation.
Bert Huang and Tony Jebara. New York Academy of Sciences Machine Learning Symposium 2007.
Poster and abstract

Maximum Entropy Density Estimation with Incomplete Data.
Bert Huang and Ansaf Salleb-Aouissi. New York Academy of Sciences Machine Learning Symposium 2007.
Poster and abstract.

Loopy Belief Propagation for Bipartite Maximum Weight b-Matching.
Bert Huang and Tony Jebara. New York Academy of Sciences Machine Learning Symposium 2006.
Poster and abstract.

Unrefereed Technical Reports

Approximating the Permanent with Belief Propagation.
Bert Huang and Tony Jebara.
pdf, arxiv.

Misc.

spouterprod.c. A useful mex utility to compute sparse outer products in matlab, taking advantage in computation time and memory usage when you are interested in computing something like mask.*(U*V'), where mask is a sparse binary matrix.