Spintronics-Based Neuromorphic and Ising Computing

With the scaling of conventional-silicon-transistor devices almost reaching saturation and conventional machine learning/neural networks implementations hitting a major bottleneck (due to the separation of memory and computing units in conventional computers, popularly known as the von Neumann bottleneck), various unconventional computing paradigms have emerged recently. Neuromorphic computing is one such brain-inspired computing paradigm which makes use of novel materials beyond silicon and devices beyond transistors to solve machine learning tasks on energy-efficient hardware for edge artificial intelligence. Magnetic materials and related nanomagnetic and spintronic devices are very
important for neuromorphic computing because of their non-volatile memory and other properties. Motivated by this, Prof Debanjan Bhowmik has carried out an extensive amount of work on spintronics-based neuromorphic computing through both experiments and simulations. He will present his major research findings in this talk, particularly focusing on spin-orbit-torque-driven non-volatile memory synapse devices based on bilayer (ferromagnet-heavy metal) and gradient multi-layer thin-film stacks. He will also discuss how inference on crossbar-array-based neuromorphic hardware can be further speeded up with the help of frequency division multiplexing.

Short biography:
Debanjan Bhowmik is currently an Associate Professor in the Department of Electrical Engineering, Indian Institute of Technology (IIT) Bombay. He obtained his PhD degree from the Department of Electrical Engineering and Computer Sciences, University of California Berkeley, in 2015, working in the field of nanomagnetism and spintronics. He worked as a faculty member in the Department of Electrical Engineering, Indian Institute of Technology (IIT) Delhi, from 2017 to 2021, and then joined IIT Bombay in January 2022. His current research interests are in implementing machine learning/optimization algorithms for edge-computing applications through emerging devices and architectures, with specific emphasis on spintronic implementation of the same.