- This event has passed.
LANS Seminar
June 1, 2022 @ 10:30 - 11:30 CDT
Seminar Title: Bayesian Brain Inspired Computing
Speaker: Maryam Parsa, Assistant Professor, Electrical and Computer Engineering Department, George Mason University
Date/Time: June 1, 2022 / 10:30 am – 11:30 am
Location: See meeting URL on the cels-seminars website (requires Argonne login)
Host: Prasanna Balprakash
Description: Neuromorphic systems allow for extremely efficient hardware implementations for neural networks (NNs). In recent years, several algorithms have been presented to train spiking NNs (SNNs) for neuromorphic hardware. However, SNNs often provide lower accuracy than their artificial NNs (ANNs) counterparts or require computationally expensive and slow training/inference methods. To close this gap, designers typically rely on reconfiguring SNNs through adjustments in the neuron/synapse model or training algorithm itself. Nevertheless, these steps incur significant design time, while still lacking the desired improvement in terms of training/inference times (latency). Designing SNNs that can mimic the accuracy of ANNs with reasonable training times is an exigent challenge in neuromorphic computing. In this talk, an alternative approach is presented that looks at such designs as an optimization problem rather than architecture design. A versatile multi-objective Bayesian-based neural architecture search (NAS) is developed for automatically tuning design metrics of several state-of-the-art SNN training algorithms, and their underlying hardware. These include evolutionary-based, conversion-based, and back propagation-based training methods on CMOS and Beyond CMOS accelerators. Without the need to redesign or invent new training algorithms/architectures, not only significant performance improvements in terms of accuracy are observed, but also the energy requirements and latency in terms of speed of training and inference are also drastically reduced. This talk will also include discussions on how Bayesian meta-learning, and Bayesian learning/inference may lead to break through results in artificial intelligence and brain-inspired computing.
Bio: Dr. Maryam Parsa is an Assistant Professor in Electrical and Computer Engineering (ECE) department at George Mason University. Prior to joining GMU, she was a Postdoctoral Researcher at Oak Ridge National Laboratory in Beyond Moore Computing group. She received her PhD in ECE from the Center for Brain Inspired Computing (C-BRIC) at Purdue university under supervision of Prof. Kaushik Roy in December 2020. She was the recipient of several prestigious awards including Intel/SRC PhD fellowship for four years, ORNL ASTRO fellowship, Purdue university Ross fellowship, and TECHCON student presenter award. She also worked at Intel Corporation in 2014 and 2016 as research intern. Dr. Parsa’s research interests are in the areas of neuromorphic computing, neural architecture search and Bayesian learning across the full stack of materials, devices, circuits, systems, algorithms, and applications.
Please note that the meeting URL for this event can be seen on the cels-seminars website, which requires an Argonne login.