Who’s Mohsen Imani
Aug 1st, 2022
Department of Computer Science,
University of California Irvine
Brain-Inspired Computing, Computer Architecture, Embedded Systems
Mohsen Imani is an Assistant Professor in the Department of Computer Science at UC Irvine. He is also a director of Bio-Inspired Architecture and Systems Laboratory (BIASLab). He is working on a wide range of practical problems in the area of brain-inspired computing, machine learning, computer architecture, and embedded systems. His research goal is to design real-time, robust, and programmable computing platforms that can natively support a wide range of learning and cognitive tasks on edge devices. Dr. Imani received his Ph.D. from the Department of Computer Science and Engineering at UC San Diego. He has a stellar record of publication with over 120 papers in top conferences/journals. His contribution has led to a new direction in brain-inspired hyperdimensional computing that enables ultra-efficient and real-time learning and cognitive support. His research was also the main initiative in opening up multiple industrial and governmental research programs. Dr. Imani’s research has been recognized with several awards, including the Bernard and Sophia Gordon Engineering Leadership Award, the Outstanding Researcher Award, and the Powell Fellowship Award. He also received the Best Doctorate Research from UCSD, the best paper award in Design Automation and Test in Europe (DATE) in 2022, and several best paper nomination awards at multiple top conferences including Design Automation Conference (DAC) in 2019 and 2020, Design Automation and Test in Europe (DATE) in 2020, and International Conference on Computer-Aided Design (ICCAD) in 2020.
Dr. Imani’s research has been instrumental in developing practical implementations of Hyper-dimensional (HD) computing – a computational technique modeled after the brain. The Hyper-dimensional computing system enabled large-scale learning in real-time, including both training and inference. He has developed such a system by not only accelerating machine learning algorithms in hardware but also redesigning the algorithms themselves using strategies that more closely model the ultimate efficient learning machine: the human brain. HD computing is motivated by the observation that the key aspects of human memory, perception, and cognition can be explained by the mathematical properties of high-dimensional spaces. It thereby models the human memory using points of a high-dimensional space, that is, with hypervectors (tens of thousand dimensions.) These points can be manipulated under a formal algebra to represent semantic relationships between objects, and thus we can devise various cognitive solutions which memorize and learn from the relation of data. HD computing also mimics several desirable properties of the human brain including robustness to noise and failure of memory cells, and one-shot learning which does not require a gradient-based algorithm. Dr. Imani exploited these key principles of brain functionalities to create cognitive platforms. The platforms include (1) novel HD algorithms supporting classification and clustering which represent the most popular categories of algorithms used regularly by professional data scientists, (2) novel HD hardware accelerators capable of up to three orders of magnitude improvement in energy efficiency relative to GPU implementations, and (3) an integrated software infrastructure that makes it easy for users to integrate HD computing as a part of systems, and that enables secure distributed learning on encrypted information using HD computing. The software contributions are backed by efficient hardware acceleration in GPU, FPGA, and processing in-memory. Dr. Imani leveraged the memory-centric nature of HD computing to develop efficient hardware/software infrastructure for a highly-parallel PIM acceleration. In HD computing, hypervectors have holographic distribution, where the information is uniformly distributed over a large number of dimensions. This makes HD computing significantly robust to the failure of an individual memory component (Robust to ∼30% failure in the hardware). In particular, Dr. Imani exploited this robustness to design an approximate in-memory associative search that checks the similarity of hypervectors in about tens of nano-seconds, while providing orders of magnitude improvement in energy efficiency as compared to today’s exact processors.