Associative memories using complex-valued Hopfield networks based on spin-torque oscillator arrays

Published

July 14, 2022

Author(s)

Nitin Prasad, Prashansa Mukim, Advait Madhavan, Mark Stiles

Abstract

Simulations of complex valued Hopfield networks based on spin-torque oscillators can recover phase-encoded images. Sequences of memristor-augmented inverters provide tunable delay elements that implement complex weights by phase shifting the oscillatory output of the oscillators. Pseudo-inverse training is sufficient to store at least twelve images in a set of 192 oscillators, representing the $16\times 12$ pixel images. The energy to recover an image depends on the desired error level. For the oscillators and circuitry considered here, 1\% root mean square deviations from the ideal image require approximately 5$\mu$s and consumes roughly 465nJ. Simulations show that the network functions well when the resonant frequency of the oscillators has a fractional spread less than $10^-3}$ depending on the strength of the feedback.

Citation

Neuromorphic Computing and Engineering

Volume

2

Issue

3

Pub Type

Journals

Download Paper

https://doi.org/10.1088/2634-4386/ac7d05

Local Download

Keywords

Spin-torque oscillators, magnetic tunnel junctions, Hopfield networks, CMOS delay circuits, memristor, image reconstruction, associative memory, spin-transfer torque, oscillators, synchronization, image reconstruction

Citation

Prasad, N. , Mukim, P. , Madhavan, A. and Stiles, M. (2022), Associative memories using complex-valued Hopfield networks based on spin-torque oscillator arrays, Neuromorphic Computing and Engineering, [online], https://doi.org/10.1088/2634-4386/ac7d05, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=933660 (Accessed October 11, 2024)