Power saving analog reservoir computing system design

Chew, Yee Yuen (2021) Power saving analog reservoir computing system design. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Preview

Text (Full Text) Power saving analog reservoir computing system design.pdf - Submitted Version Download (3MB) | Preview

Abstract

The aim of this project is to design an Analog power saving Spike-Based Delayed Feedback Reservoir Computing System and analyse its performance. The developed system can make the integrated circuit (IC) achieves power saving so that can solve the problem excessive power consumption of supercomputer with good performance. The rate of augmentation is starting to slow down due to the underlying performance limits of the chips, indicating the end of Moore's prediction. Moore's prediction is coming to an end as things are starting to slow down. The urge to get through the barrier has led researchers down a few paths, including innovative computing architectures. Over the last few years, reservoir computing has evolved as a revolutionary notion in the field of machine learning. To process temporal data, reservoir computing combines the memory and Spatio-temporal processing capabilities of recurrent neural networks. In this work, a new class of computationally efficient spike timing-dependent encoders and delay-based reservoirs within reservoir networks has been proposed. Silterra 130nm technology is used to develop the reservoir computing system. The proposed method eliminates the need for power hungry analog-to-digital converters (ADCs) and operational amplifiers (Op-AMPs), resulting in lower power consumption and a smaller design footprint.

Actions (login required)

View Item