Acceleration of Deep Neural Network Training with Resistive Cross-Point Devices

I fully expect similar approaches to come to dominate artificial neural network implementations, due to their ability to accelerate on-line learning. Either memristive or ferroelectric (FRAM) nonvolitile memory could be used to implement this technology in the very near future; note that TI has FRAM in commercially available chips.

I suspect some of the bigger players have research along these lines, but I don't really know. If there are any (well funded) groups onboarding in this field, let me know, though. I have extensive experience in R&D at the basic semiconductor materials, device and systems levels, and I've studied the software techniques on the side.