Title:  A Biologically Plausible Sparse Approximation Solver on Neuromorphic Hardware

Committee:

Dr. David Anderson, ECE, Chair , Advisor

Dr. Justin Romberg, ECE

Dr. Christopher Rozell, ECE

Dr. Mark Davenport, ECE

Dr. Andreas Andreou, Johns Hopkins

Abstract:

This work delivers a neuromorphic system that solves for the sparse approximation on hardware geared toward real-world embedded systems signal processing applications.  We choose to explore the biologically-plausible Locally Competitive Algorithm (LCA), a neural network that solves the sparse approximation problem.  We implement this algorithm on the brain-inspired IBM Neurosynaptic System, also known as the TrueNorth chip, a specialized hardware platform that has shown success in deploying neural networks for signal processing applications.