In this talk I will introduce a novel, brain inspired from of computation, which encodes information in time, instead of encoding information as voltage levels as is done in classical computing. Dubbed “Race Logic”, such an encoding strategy allows some simple yet powerful operations to become trivial to implement with today’s CMOS and emerging technology based systems. In such a scheme, computations are performed by observing the relative propagation times of signals injected into a configurable circuit (i.e. the outcome of races through the circuit).
I will begin by introducing different ways in which the human endeavor of ever increasing compute performance and energy efficiency has historically been guided by attempts to understand the brain. I will describe the challenges of performing biology like computation with non-biological elements and show how a synergistic combination of encoding schemes, architectures, circuits and emerging technologies can be used to improve throughput per unit energy by 4 to 5 orders of magnitude over classical computation techniques for certain problems. I will also present future directions of such a computational paradigm and touch upon how it could be used to solve classification based problems.
University of California, Santa Barbara