Professor Incorvia collaborates with Patrick Xiao and Nicholas Zogbi on an article about using domain wall-magnetic tunnel junctions for parallel matrix multiplication. They solve a couple major challenges in using magnetic domain walls and magnetic tunnel junctions for computing by using voltage to increase the device reliability, and device design to allow multiple fanout without change in device footprint. With these changes, they are able to use physics-based simulations to show a fully functioning 8-bit MAC operation, the main computation for deep neural network inference!
Click here to read the article!
