Your browser is unsupported

We recommend using the latest version of IE11, Edge, Chrome, Firefox or Safari.

Feb 2 2024

Stochastic Computing 2.0: Deterministic, Accurate, Fast, Robust, and Light Bit-Stream Processing for Energy-Efficient AI

ECE 595 Seminar Series

February 2, 2024

11:00 AM - 12:15 PM


Lecture Center C3


801 S. Halsted, Chicago, IL 60607

Stochastic Computing 2.0: Deterministic, Accurate, Fast, Robust, and Light Bit-Stream Processing for Energy-Efficient AI

Presenter: M. Hassan Najafi, University of Louisiana, Lafayette

Abstract: Despite the commercial success of machine learning and artificial intelligence, radically novel computing paradigms must be explored if we ever hope to achieve brain-like capabilities with brain-like efficiencies. One potential class of paradigms, consistent with important aspects of neocortical operations, is stochastic (aka unary) computing (SC). SC uses a simple number representation to achieve low-cost and robust implementations for complex arithmetic functions. SC-based designs consistently achieve 50x to 100x reductions in gate count and significantly higher tolerance to soft-error compared to conventional binary implementations. The paradigm also offers fully parallel and scalable hardware implementations. Despite all its positives, the conventional SC (v1.0) suffers from high conversion cost, inaccuracy, and long latency. In this talk, I will present our recent advances in this re-emerging area and introduce SC 2.0, which is deterministic, accurate, fast, robust, low-cost, and most importantly, energy efficient. The proposed computational model is an excellent fit for designing cost-efficient and ultra-low-power ML and AI systems. It enables the next generation of data-intensive applications with computing embedded in data, creating intelligent and robust in-memory architectures that do as much computing as possible close to the bits. We further discuss a radically novel and highly unorthodox idea for fast, energy-efficient, and seamless data processing near sensors. Finally, as the first study of its kind, we discuss the integration of SC and brain-inspired hyperdimensional computing (HDC) to design highly efficient learning systems.

Speaker bio: M. Hassan Najafi is an assistant professor in the School of Computing and Informatics at the University of Louisiana at Lafayette, where he directs the Emerging Computing Technologies (ECT) Research Lab. He is working on a wide range of practical problems in brain-inspired computing, stochastic computing, in-memory computing, near-sensor processing, hyperdimensional computing, computer architecture, machine learning, and ultra-low-power VLSI design. Najafi received his MS degree in computer architecture from the University of Tehran, Iran, in 2013 and his PhD in electrical engineering from the ECE department of the University of Minnesota-Twin Cities in 2018. He has authored/co-authored more than 80 peer-reviewed papers in top conferences/journals and has been granted five U.S. patents with more pending. Najafi’s research has been recognized with several awards, including the Best Paper Award in the 33rd Great Lakes Symposium on VLSI (GLSVLSI'23), the 2023 Lockheed Martin Corporation Endowed Professorship in Computer Science and Engineering, the 2022 Francis Patrick Clark/BORSF Professorship in Computing and Informatics, the Best Research Award at the 2019 DAC PhD Forum, the 2018 EDAA Outstanding Dissertation Award, the Best Paper Award at the IEEE International Conference on Computer Design (ICCD’17), and the Feature Paper of the Month in IEEE Transactions on Computers. Najafi has served as a panelist for multiple U.S. federal agencies, including NSF and DoE, an editor for the IEEE Journal on Emerging and Selected Topics in Circuits and Systems, and a technical committee member and reviewer for more than 55 top IEEE and ACM conferences and journals.

Faculty host: Natasha Devroye,


ECE department

Date posted

Jan 30, 2024

Date updated

Jan 30, 2024