Single-pattern filter algorithm for training analog artificial neural networks: a thesis in Electrical Engineering

Haval Marsho Elias

doi:10.62791/20220

Back

Thesis

Open access

Single-pattern filter algorithm for training analog artificial neural networks: a thesis in Electrical Engineering

Haval Marsho Elias

Master of Science (MS), University of Massachusetts Dartmouth

2022

DOI:

https://doi.org/10.62791/20220

Abstract

Artificial Neural Networks (ANNs) are typically realized as computer programs designed to simulate the way in which the human brain processes information. ANNs gather their knowledge by detecting the patterns and relationships through training not from programming. This thesis is using (SMART) “Self-trained Multi-layer Analog Real-time” architecture to solve the handwritten pattern recognition problem for numbers (0-9) by introducing a “Single-Pattern Filter Algorithm” (SPF) which takes inputs from raw data in the MNIST file for handwritten numbers (0-9) and provides 10×10 outputs to set the weights for the 10 neurons in the (OLNs) “output layer neurons”. The SPF algorithm is implemented in MATLAB.

Files and links (1)

pdf

Elias H.M. COE MS Thesis 2022786.69 kBDownload View

CC BY-NC-ND V4.0, Open Access

Metrics

2 File views/ downloads

10 Record Views

Details

Title: Single-pattern filter algorithm for training analog artificial neural networks
Creators: Haval Marsho Elias
ORCID: 0000-0002-1690-7219
Contributors: David P Rancour (Advisor) - University of Massachusetts Dartmouth, Department of Electrical and Computer Engineering
Dayalan P. Kasilingam (Committee Member) - University of Massachusetts Dartmouth, Department of Electrical and Computer Engineering
Paul J Fortier (Committee Member) - University of Massachusetts Dartmouth, Department of Electrical and Computer Engineering
Number of pages: ix, 33 pages
Illustrations: illustrations (some color)
Table of contents: List of figures -- List of tables -- Abbreviations -- Chapter 1. Introduction -- An introduction to artificial neural networks -- Types of neural networks -- Definition of the terms used -- Literature review -- Chapter 2. Methods used -- Requirements and specifications -- Network design -- Single-pattern filter algorithm -- Description of SPF algorithm -- Outliers in neural networks -- Chapter 3. Results and discussion -- MATLAB calculations for MNIST database -- Neural networks design -- Description of the program used -- Conclusion and discussion -- Future scope -- References -- Appendix: MATLAB coding.
References: Includes bibliographical references (page 29).
Awarding Institution: University of Massachusetts Dartmouth
Degree Awarded: Master of Science (MS)
Degree in: Electrical Engineering
Academic Unit: Department of Civil and Environmental Engineering
Language: English
Resource Type: Thesis
DOI: https://doi.org/10.62791/20220
Record Identifier: 9914424800001301