Title: Overcoming Noise and Variations in Low-precision Neural Networks
Dr. David Anderson, ECE, Chair , Advisor
Dr. Arijit Raychowdhury, ECE
Dr. Aaron Lanterman, ECE
Dr. Shaolan Li, ECE
Dr. Hyesoon Kim, CoC
Abstract: Traditional machine learning algorithms and neural networks are implemented using powerful digital computational architectures such as GPUs, TPUs, and FPGAs, demonstrating high performance and successfully completing previously impossible tasks. Unfortunately, the power required to train and generate predictions with the neural networks is too high to be implemented in energy-constrained systems such as implants and edge devices. Many of these systems would significantly benefit from on-board neural networks that could respond to stimuli in real time. The important question that this work seeks to address is how to bring the game-changing power of neural networks closer to the edge of the internet of things without significant degradation of performance or battery life.