Memristive Stochastic Computing for Deep Learning Parameter Optimization

Stochastic Computing (SC) is a computing paradigm that allows for the low-cost and low-power computation of various arithmetic operations using stochastic bit streams digital logic. In contrast to conventional representation schemes used within binary domain, sequence in domain inconsequential, usually non-deterministic. this brief, we exploit stochasticity during switching probabilistic Conductive Bridging RAM (CBRAM) devices efficiently generate order perform Deep Learning (DL) parameter optimization, reducing size Multiply Accumulate (MAC) units by 5 orders magnitude. We demonstrate 40-nm Complementary Metal Oxide Semiconductor (CMOS) process our scalable architecture occupies 1.55mm$^2$ consumes approximately 167$\mu$W when optimizing parameters Convolutional Neural Network (CNN) while it being trained character recognition task, observing no notable reduction accuracy post-training.