Relational Reasoning and Cortical Computation
How does thinking work? How does the cortex compute? This is one of today's greatest mysteries in science. We do not yet know how to make machines do computations similar to the computations done with ease by animal brains. By experimenting with cortically inspired architectures, we hope to gain an understanding of how such computation can occur.
One of our current directions is examining models similar to belief propagation on factor graphs, and how such models can be adapted to come closer to the style of computation found in the cortex, while solving learning and control problems of the sort that brains solve naturally.