WBT/CEA-Leti Conference Paper on Neuromorphic Computing

Hi all, long time lurker, first time poster. Was wondering if anyone had access to the full document just put up on the web today here: https://ieeexplore.ieee.org/abstract/document/9073840/authors#authors

It's a paper in progress Authored by Amir Regev and a handful of other WBT/Leti researchers, slated to be presented at an IEE international conference on Artificial Intelligence Circuits and Systems (AICAS) in late August/early September (so it hasn't been peer-reviewed yet). The Abstracts TL;DR is that they're achieving a classification accuracy of handwritten digits of 82%. This number as a whole isn't great, as we can achieve upwards of 99% accuracies with Deep CNN's (convolutional neural networks) at the moment, however, the fact that it was achieved using WBT's ReRAM in a physical SNN is big news! If this is their first sort of test case, then it's only upwards from here accuracy wise.

This all could be a huge proof of concept to potential Neuromorphic Computing partners, so if anyone has access to the rest of that paper please let me know as I'd love to do some further reading!