We propose a mechanism for binding and automatic inference in humans that does not rely on temporal synchrony. An online example and demo is available.

We give an algorithm for learning all optimal policies when performing reinforcement learning with multiple rewards.

We propose a mechanism for binding and automatic inference in humans that does not rely on temporal synchrony. An online example and demo is available.

Automatically learning more informative grammars gives grammars that capture linguistic features and also gives high parsing accuracy; we got an F1 score of 90.2.

We examine parsing text by using a hierarchical mixture of grammars. This helps capture non-local correlations, but unfortunately these make up a small amount of all correlations to be captured.

My first published paper! Automatic camera placement and multiple-camera interpolation, with Cindy Grimm and others.

A report on how neural computation is being used in the "real world" by companies; my class project for VS298 (Neural Computation).

Slav's and my class project for Wilensky's Natural Language class

Alex's and my class project for Papadimitriou's Theory class

My senior thesis at Washington University in St. Louis. I used a computer model of a crystal to help interpret X-ray diffraction scans.

My Computer Vision class project. I analyzed a time-lapse video and tried to correct for lighting changes.