Abstract
Neural networks with memristive memory for weights have been proposed as an energy-efficient solution for scaling up of neural network implementations. However, training such memristive neural networks is still challenging due to various memristor imperfections and faulty memristive elements. Such imperfections and faults are becoming increasingly severe as the density of memristor arrays increases in order to scale up weight memory. We propose fault pruning, a robust training scheme for memristive neural networks based on the idea to identify faulty memristive behavior on the fly during training and prune corresponding connections. We test this algorithm in simulations of memristive neural networks using both feed-forward and convolutional architectures on standard object recognition data sets. We show its ability to mitigate the detrimental effect of memristor faults on network training.
Original language | English |
---|---|
Title of host publication | Unconventional Computation and Natural Computation, UCNC 2023 |
Editors | Daniela Genova, Jarkko Kari |
Place of Publication | Cham |
Publisher | Springer |
Pages | 124-139 |
Number of pages | 16 |
ISBN (Electronic) | 978-3-031-34034-5 |
ISBN (Print) | 978-3-031-34033-8 |
DOIs | |
Publication status | Published - 2023 |
Event | 20th International Conference on Unconventional Computation and Natural Computation - University of North Florida in Jacksonville, Jacksonville, United States Duration: 13 Mar 2023 → 17 Mar 2023 https://sites.google.com/view/ucnc2023/ |
Publication series
Name | Lecture Notes in Computer Science |
---|---|
Volume | 14003 |
Conference
Conference | 20th International Conference on Unconventional Computation and Natural Computation |
---|---|
Abbreviated title | UCNC 2023 |
Country/Territory | United States |
City | Jacksonville |
Period | 13/03/23 → 17/03/23 |
Internet address |
Keywords
- neural networks
- memristors
- robust training
- memristor faults
- network pruning
- Neural networks
- Robust training
- Memristors
- Network pruning
- Memristor faults
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)