Convolution Inference via Synchronization of a Coupled CMOS Oscillator Array

Oscillator neural networks (ONN) are a promising hardware option for artificial intelligence. With an abundance of theoretical treatments of ONNs, few experimental implementations exist to date. In contrast to prior publications of only building block functionality, we report a practical experimental demonstration of neural computing using an ONN. The arrays contain 26 CMOS ring oscillators in the GHz range of frequencies tuned by image data and filters. Synchronization of oscillators results in an analog output voltage approximating convolution neural network operation.

[1]  Eugene M. Izhikevich,et al.  Class 1 neural excitability, conventional synapses, weakly connected networks, and mathematical foundations of pulse-coupled models , 1999, IEEE Trans. Neural Networks.

[2]  S. Mudanai,et al.  22FFL: A high performance and ultra low power FinFET technology for mobile and RF applications , 2017, 2017 IEEE International Electron Devices Meeting (IEDM).

[3]  Wolfgang Porod,et al.  Computing With Networks of Oscillatory Dynamical Systems , 2019, Proceedings of the IEEE.

[4]  Renyuan Zhang,et al.  CMOS supporting circuitries for nano-oscillator-based associative memories , 2012, 2012 13th International Workshop on Cellular Nanoscale Networks and their Applications.

[5]  Damien Querlioz,et al.  Vowel recognition with four coupled spin-torque nano-oscillators , 2017, Nature.

[6]  Tadashi Shibata,et al.  Coupled-Oscillator Associative Memory Array Operation for Pattern Recognition , 2015, IEEE Journal on Exploratory Solid-State Computational Devices and Circuits.

[7]  Dmitri E. Nikonov,et al.  Benchmarking Physical Performance of Neural Inference Circuits , 2019, ArXiv.