A General Purpose Neural Processor

Thumbnail Image

Files

Mountain_umbc_0434D_11604.pdf (2.68 MB)

Links to Files

Permanent Link

http://hdl.handle.net/11603/15519

Collections

UMBC Theses and Dissertations

Author/Creator

Author/Creator ORCID

Date

2017-01-01

Type of Work

dissertations
Text

Department

Computer Science and Electrical Engineering

Program

Engineering, Computer

Citation of Original Publication

Rights

This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu
Distribution Rights granted to UMBC by the author.

Subjects

Computer Architecture
Cybersecurity
Memristors
Neural Nets
Neuromorphic Computing

Abstract

Computer applications are evolving from traditional scientific and numerical calculations, to a more diverse set of uses including speech recognition, robotics, and analytics. This has created a fertile environment for the investigation of non-traditional programming approaches and models of computing, inspired by neuroscience, often termed neuromorphic computing. Neural nets have emerged as one of the primary neuromorphic computing approaches; von Neumann architectures, conceived for scientific computing applications are not optimized for neural nets. This research focuses on developing a general purpose computer architecture optimized for neural net based applications. The architecture is useful for a variety of learning algorithms, and is evaluated across a spectrum of potential applications. Both traditional and emerging technologies are explored, with trade-offs being made based on the most important system level metrics.