Abstract
In this work, online training is experimentally demonstrated on a neuromorphic network built with 1k-bit 1T1R RRAM array. The 1T1R RRAM cells in the array exhibit excellent synaptic behavior. Patterns with up to 11.83% noises can be correctly classified by the network after training. Based on the analysis of experimental results, we find the device characteristics and operation schemes significantly affect the system performance. The impact of voltage dependence and asymmetry effects are evaluated, and optimization guidelines are provided.
Original language | English (US) |
---|---|
Title of host publication | 2017 IEEE Electron Devices Technology and Manufacturing Conference, EDTM 2017 - Proceedings |
Publisher | Institute of Electrical and Electronics Engineers Inc. |
Pages | 182-183 |
Number of pages | 2 |
ISBN (Electronic) | 9781509046591 |
DOIs | |
State | Published - Jun 13 2017 |
Event | 2017 IEEE Electron Devices Technology and Manufacturing Conference, EDTM 2017 - Toyama, Japan Duration: Feb 28 2017 → Mar 2 2017 |
Other
Other | 2017 IEEE Electron Devices Technology and Manufacturing Conference, EDTM 2017 |
---|---|
Country/Territory | Japan |
City | Toyama |
Period | 2/28/17 → 3/2/17 |
Keywords
- network
- RRAM
- synaptic behavior
ASJC Scopus subject areas
- Hardware and Architecture
- Electrical and Electronic Engineering