Implementation of quantized artificial neural networks with spintronic stochastic computing

Author

Saadi Sabyasachi Mr., Virginia Commonwealth UniversityFollow

Author ORCID Identifier

https://orcid.org/0009-0008-2212-8992

Defense Date

2025

Document Type

Thesis

Degree Name

Master of Science

Department

Mechanical and Nuclear Engineering

First Advisor

Jayasimha Atulasimha, Ph.D.

Abstract

Artificial intelligence or machine learning is going through a rapid expansion. It also incurs significant costs for power and device footprints. Various approaches are being explored to design energy and hardware efficient machine learning models. Stochastic computing has been proposed for efficient machine learning implementation. It requires a source of random number generation which poses some practical challenges. So spintronic solutions such as magnetic tunnel junction has been used for random number generation. Again, spintronic random number generation to implement high precision circuit is prone to device-to-device variations. Hence we designed quantized artificial neural network with spintronic stochastic computing which is shown to retain accuracy with lower energy consumption and latency and also resilient to device to device variations.

While we applied this approach for simple one hidden layer and three hidden layer architectures of deep neural networks, it can be extended to convolutional neural network which would enable it to be applicable for more complex image dataset.

Another perspective of spintronic application is spatiotemporal data classification with reservoir computing. Spatiotemporal data has a time dimension along with spatial dimension. A concurrent approach for spatiotemporal data analysis is convolutional neural network with long short term memory (CNN-LSTM). Spintronic systems have inherent memory property which allows to do memory based computations under certain limitations which can significantly improve energy efficiency for edge applications.

Overall, thesis Master’s thesis covers spintronic quantized neural network for spatial data classification and explores spintronic reservoir based model for spatiotemporal data classification.

Rights

© The Author

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission

12-15-2025