
A spin-orbit torque (SOT) magnetoresistive random-access memory (MRAM)-based probabilistic binary neural network (PBNN) system that achieves high classification accuracy, noise-tolerance, and resource-saving through the use of controllable random weight matrices and a compute-in-memory architecture.


coremsg

noise-tolerant-and-resource-efficient-probabilistic-binary-neural-network-implemented-with-sot-mram-compute-in-memory-system


Noise-Tolerant and Resource-Efficient Probabilistic Binary Neural Network Implemented with SOT-MRAM Compute-in-Memory System
