Andrew E. Gelfand

My primary focus is on inference and learning in graphical models. I am interested in developing methods to efficiently perform inference in probabilistic models containing deterministic relations (constraints) and in developing approximate inference methods that exploit problem structure. In particular, I am interested in understanding how the choice of regions (i.e. clusters, cliques) impact the convergence and accuracy of the class of (Generalized) Belief Propagation message passing algorithms and in developing methods to automate the selection of regions in general Markov networks.

Below are a list of publications and technical reports.

Conferences

• Generalized Belief Propagation on Tree Robust Structured Region Graphs    [Proofs]
    Andrew E. Gelfand and Max Welling
    In 28th Conference on Uncertainty in Artificial Intelligence (UAI). Catalina Island, CA, USA, August 2012.
• A Cluster-Cumulant Expansion at the Fixed Points of Belief Propagation
    Max Welling, Andrew E. Gelfand and Alexander Ihler
    In 28th Conference on Uncertainty in Artificial Intelligence (UAI). Catalina Island, CA, USA, August 2012.
• Integrating Local Classifiers through Nonlinear Dynamics on Label Graphs with an Application to Image Segmentation
    Yutian Chen, Andrew E. Gelfand, Charless Fowlkes and Max Welling
    13th IEEE International Conference on Computer Vision (ICCV) Barcelona, Spain, November 2011.
• Stopping Rules for Randomized Greedy Triangulation Schemes
    Andrew E. Gelfand, Kalev Kask and Rina Dechter
    AAAI 2011. San Francisco, CA, USA, August 2011.
• Pushing the Power of Stochastic Greedy Ordering Schemes for Inference in Graphical Models
    Kalev Kask, Andrew E. Gelfand, Lars Otten, and Rina Dechter
    AAAI 2011. San Francisco, CA, USA, August 2011.
• On Herding and the Perceptron Cycling Theorem
    Andrew E. Gelfand, Laurens van der Maaten, Yutian Chen and Max Welling
    NIPS 2010. Vancouver, Canada, December 2010.
• BEEM : Bucket Elimination with External Memory
    Kalev Kask, Rina Dechter and Andrew E. Gelfand.
    In 26th Conference on Uncertainty in Artificial Intelligence (UAI). Catalina Island, CA, USA, July 2010.
• Performance Evaluation of Decentralized Estimation Systems with Uncertain Communication
    Andrew E. Gelfand, Chris Smith, Mike Colony and Christopher Bowman
    12th International Conference on Information Fusion. Seattle, WA, USA, July 2009.

Technical Reports

• A Comparison of Algorithms for Collaborative Filtering on RBMs
    Andrew E. Gelfand, March 2010
• Post-Processing Elimination Orderings to Reduce Induced Width
    Andrew E. Gelfand, December 2009
• Solving Cryptograms with the Constrained Cyrpto-EM Algorithm
    Andrew E. Gelfand, December 2009

My (hopefully) up-to-date Resume