Artificial Synapse with Mnemonic Functionality using GSST-based Photonic Integrated Memory

Abstract

Here we present a multi-level discrete-state nonvolatile photonic memory based on an ultra-compact ( ) hybrid phase change material GSST-silicon Mach Zehnder modulator, with low insertion losses (3dB), to serve as node in a photonic neural network. Emulating an opportunely trained 100×100 fully connected multilayered perceptron neural network with this weighting functionality embedded as photonic memory, shows up to 92% inference accuracy and robustness towards noise when performing predictions of unseen data.