2020 BRN Discussion, page-314

Also on the same playbill:
Spiking neural networks enabling massively parallel, low-power & low-latency computation
PRESENTER: Bruno PAING
VP Strategic partnerships, North America LETI
TYPES OF ITEMS TO BE DEMONSTRATED:
CEA-Leti will introduce ‘SPIRIT’, the world-first fully integrated neural network on-chip with nonvolatile resistive memory. So far, memories were placed outside of chips leading to high energy
consumption. With this co- integration in the same die of analog spiking neurons and resistive
synapses leveraging resistive random access memory cells (RRAM), CEA-Leti enables the push for
distributed computing devices to support artificial intelligence at the edge. These spiking neural
networks are designed by CEA-Leti and the RRAM are fabricated in a post-processat CEA-Leti on
CMOS-based wafers.
RELEVANCE TO TINYML:
SPIRIT allows massively parallel, low-power and low latency computation. It is a perfect candidate for
embedded classification applications required in Lidars or IoT devices. The technology also finds
applications in event-based sensors packed with rich temporal content

This uses ReRAM their US787660 (dating from 2004) uses co-linear phase change current to change a material fron crystalline to amorphous to change resistance. Perviously done by transverse current.

ReRAM has come up for discussion before but this is the first application I have seen (not that I've been looking).

I think Tech or one of our other champoin posters may have discussed ReRAM with PvdM.