Institute Member Leslie M. Kay, Ph.D., with colleagues from Cornell University and University of Texas at Austin, received funding for her proposal titled  “Context-Dependent Reconfiguration of an Intelligent Neural System.”

The Defense Advances Research Projects Agency (DARPA) Microsystems Technology Office selected the proposal for their Lifelong Learning Machines program.

Kay and her colleagues will produce a neural network-based system that can adapt to multiple tasks and contexts. It will adjust strategies based on external context, such as an opponent’s strategy or a change in task constraints or difficulty, and to multiple internal contexts, such as prior learning or energy availability. We will build models based on the rat olfactory system, a system that accomplishes all of the above feats. Our experiments will discover how higher order areas in the brain manipulate low level input layers based on expectation, attention, and reward, and how these functions change depending on energy levels (glucose availability) across multiple time scales.