Focused Surface Acoustic Wave Induced Nano-oscillator Based Reservoir Computing

We show by micromagnetic simulations that a nanomagnet (NM) array excited by surface acoustic waves (SAW) can work as a reservoir that exhibits high short-term memory and parity check capacities. It is also able to classify sine and square waves. The simulated array has an input NM that is excited with a 4 GHz focused SAW, and 7 output NMs. The magnetizations of the output nanomagnets are processed by reading the reservoir state every 1 ns while the period of the input signal is 10 ns. The envelopes of the output NMs’ magnetization are used to train the output weights using regression method [1, 2]. For classification, a random sequence of 100 square and sine wave samples are used, of which 80 % are trained, and 20 % used for testing. We achieve 100% training and testing accuracy for different combination of NMs as outputs. Moreover, the STM and PC are calculated to be 5.5 bits and 5.3 bits which is indicative of the proposed NM array being well suited for physical reservoir computing applications [3].

1. A. J. Edwards et. al., arXiv preprint arXiv:2103.09353 (2021).

2. J. Torrejon, J. et. al., Nature, 547(7664), pp.428-431 (2017).

3. G. Tanaka et. al., Neural Networks, 115, pp.100-123 (2019).