projects

2009-present

Since 2009, I have been working full-time on general purpose machine learning algorithms. Currently, the only useful, general-purpose methods (e.g. Boltzmann machines) are quite problematic: they involve too many free parameters, require fine-tuning for different data sets, make strong hidden assumptions about the data distribution, etc. I am looking for a better theoretical foundatation for inference over high-dimensional data distributions which removes much of the arbitrariness of current methods but still provides practical algorithms.

2 0 0 8	Brainpower Labs, a software company	brain-inspired artificial general intelligence
	hierarchical topographic mixture model	Netflix Prize machine learning contest
	iBonsai: stylized 3D randomly-generated trees	simple C++ loop parallelization

2 0 0 7	Somatotopic map experiments			Simulation of a cellular automata physics model
2 0 0 7	Physically realistic fish behavior simulation			Interactive Conway's Game of Life

2 0 0 6	Experiments with branching systems	A virtual tropical island
	A computational model of the cerebellum	Self-organizing maps
	A simple C++ runtime performance profiler	Magnetic toy simulation

2 0 0 5	Measuring the benefits of curiosity-driven exploration			An open source speech i/o library
2 0 0 5	A simple C++ unit testing framework			An experimental menu system for VR applications

2 0 0 4	A C++ library for general purpose reinforcement learning	Immersive command & control in virtual reality
	Rhino: A 3D space shooter	OPAL: A C++ physics simulation library with an intuitive API
	Flammable, destructable cabin simulation	Basic ray-tracing experiments

2 0 0 3	Simulated virtual people learn to stand, jump, and walk	Interactive 3D medieval environment with realistic physics
	RTS game of backyard strategic warfare	3D animation of a morning scene
	GUI and visualization tool for evolving simulated grid-based robot behavior	Reconfigurable virtual military humvee
	Jazz quartet