Georgia Institute of Technology
Neuromorphic Approaches for Asynchronous Computing and Sensing
Abstract
dc:description.abstractThe remarkable proficiency of the human nervous system in sensory data handling is largely due its ability to process asynchronously under the governance of complicated, nonlinear differential equations. This work aims to address three major questions pertaining to the design of neuromorphic information processing systems: (1) How to efficiently simulate physically implementable biorealistic neuron models on a digital platform? (2) How to implement networks of biorealistic neurons on a general-purpose analog computing platform? (3) How to efficiently digitize the solution of neuronal ordinary differential equations (ODEs) obtained from analog platforms? These three seemingly innocuous questions are currently some of the most significant obstacles to the development of neuromorphic systems with a similar magnitude of performance as the human brain. This work begins by providing background motivation for each of these questions. We then discuss previous software-based neuron models. The introductory text concludes with an overview of transistor-channel modeling for biological circuits. Chapter 4-6 represent the main technical contribution of this work, with one chapter dedicated to the study of each of these questions. Each chapter elucidates our proposed solutions, providing relevant comparisons to previous approaches; findings are used to motivate subsequent chapters and future work.
Degree
thesis:*- Level thesis:degree_level
- Masters
- Department dc:contributor.department
- Electrical and Computer Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Bhattacharyya, Swagat
- Advisor dc:contributor.advisor
-
- Hasler, Jennifer
- Committee members dc:contributor.committeemember
-
- Sathe, Visvesh S.
- Lanterman, Aaron D.
Subjects
dc:subject × 4Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/75588