 |
Center for Nonlinear Studies |
Building a brain with electronics is hard both algorithmically and from a power density perspective. This talk will focus on three key areas: (1) Math--nonlinear dynamics, tensor analysis, bias variance dilemma. (2) Hardware--Rent's rule for the brain, CV^2f losses, spatial vs. temporal locality, memristive memory, photonic communication, analog vs. digital. (3) Algorithms--time, space, frequency domains, steerable filter theory, FFTs, tensor convolution, examples. The talk will include demonstrations of efficiently solving some essential low-level vision processes (diffusion, contrast normalization, boundary completion) on our platform, Cog ex Machina.
Host: Information Science and Technology Center (ISTC), Garrett Kenyon, gkenyon@lanl.gov, 667-1900