Publication Date

12-2021

Date of Final Oral Examination (Defense)

10-5-2021

Type of Culminating Activity

Dissertation

Degree Title

Doctor of Philosophy in Electrical and Computer Engineering

Department

Electrical and Computer Engineering

Major Advisor

Kurtis D. Cantley, Ph.D.

Advisor

Nader Rafla, Ph.D.

Advisor

Kris Campbell, Ph.D.

Advisor

Sin Ming Loo, Ph.D.

Abstract

In this dissertation, neuromorphic circuits are used to implement spiking neural networks in order to detect spatiotemporal patterns. Unsupervised training and detection-by-design techniques were used to attain the appropriate connectomes and perform pattern detection.

Unsupervised training was performed by feeding random digital spikes with a repeating embedded spatiotemporal pattern to a spiking neural network composed of leaky integrate-and-fire neurons and memristor-R(t) element circuits which implement spike-timing-dependent plasticity learning rules.

Detection-by-design was achieved using neuromporphic circuits and digital logic gates. When detection-by-design was achieved using both neuromorphic circuits and digital logic gates, a network was created of spatiotemporal pattern detector circuits, each of which was capable of detecting the three fundamental spatiotemporal patterns (NA-NA-Δt, NA-NB-Δt, and NA-NB-Coincidence), in order to detect combinations of two-spike features in the desired spatiotemporal pattern. The spatiotemporal pattern was detected when all of the two-spike features were detected. Similarly, when detection-by-design was achieved using only neuromorphic circuits, a Complex Pattern Detecting Network was was formed by combining Simple Pattern Detecting Networks, each of which was capable of detecting the three fundamental spatiotemporal patterns. The Complex Pattern Detector was used in a proof-of-concept to demonstrate a detect-and-generate spatiotemporal symbol computing paradigm.

DOI

https://doi.org/10.18122/td/1887/boisestate

Download

Share

COinS