Memcapacitive Reservoir Computing
Published In
Proceedings of the IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH 2017)
Document Type
Citation
Publication Date
2017
Abstract
Memristors have successfully been used to build efficient reservoir computers. The power consumption of memristive reservoirs, however, is bounded by the resistive nature of such devices. Here, we show that memcapacitors, another device in the mem-device family, offer great promise for power-efficient reservoir computers.We simulated memcapacitive reservoirs with two different device models and benchmarked them with the NARMA-30 and the MNIST task. The results were compared to two memristive reservoirs as well as to a software echo state network. The memcapacitive reservoirs achieved comparable performance as the memcapacitive reservoirs but reduced the power consumption by about a factor of 500× for both tasks. We argue that memcapacitive reservoirs thus have great potential for low-power neuromorphic applications.
Persistent Identifier
https://archives.pdx.edu/ds/psu/25933
Publisher
IEEE
Citation Details
Tran, S. D., & Teuscher, C. (2017, July). Memcapacitive reservoir computing. In 2017 IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH) (pp. 115-116). IEEE.