Title: Simulating Spiking Neural Networks with SuperNeuroMAT for Energy-Efficient AI
Abstract: Neuromorphic computing, which emulates the brain’s spiking dynamics, offers a path to AI that uses far less power than conventional CPUs and GPUs. Neuromorphic hardware is still hard to access, though, and the available simulators are either tied to neuroscience or deep learning workloads or too slow to scale. This talk introduces SuperNeuro, a fast and scalable simulator for neuromorphic computing, with a focus on SuperNeuroMAT, its matrix-based mode. SuperNeuroMAT expresses neuron updates as vector operations and synaptic operations, including delays and spike-timing-dependent plasticity (STDP) learning, as matrix operations. This supports sparse and dense execution with GPU acceleration. I will cover the building blocks of neuromorphic computation and prior work on its theory and applications, then present SuperNeuroMAT benchmarks against existing simulators. Examples include spiking neural networks for event-based vision (N-CARS, ASL-DVS), spiking graph neural networks, and neuromorphic algorithms. I will close with ongoing efforts to scale SuperNeuroMAT to HPC systems and to build end-to-end spiking AI models, including spiking transformers.
Bio: Dr. Prasanna Date is a Senior Scientist in the Computer Science and Mathematics Division at Oak Ridge National Laboratory, where his research applies AI and machine learning to non-conventional computing platforms, namely neuromorphic and quantum computing. His work has been recognized with two R&D 100 Awards (SuperNeuro in 2023 and MAQ in 2024) and a place on the 2022 Forbes 30 Under 30 Asia list. He has authored over 70 publications and serves as Associate Editor of IEEE Transactions on Neural Networks and Learning Systems. He has also served as General Co-Chair of the ACM International Conference on Neuromorphic Systems (ICONS) since 2024, and he is an IEEE Senior Member. He leads DOE-funded research, including the Genesis Mission project on spiking transformers. Over the past several years, he has mentored dozens of students and early-career researchers, from high school interns to postdoctoral researchers.