Each year, through the Young Faculty Award (YFA) program, DARPA identifies “rising stars” in research positions, supporting their work in the context of national security needs. This year, DARPA selected UC Irvine’s Assistant Professor of Computer Science Mohsen Imani as one of those rising stars.
“This accolade is bestowed upon only around 20 exceptional emerging talents,” says Tony Givargis, chair of the Department of Computer Science in the Donald Bren School of Information and Computer Sciences (ICS). “In particular, Mohsen has been granted $1 million to spearhead research on hyperdimensional computing as a neuro-symbolic AI technique, aiming to facilitate transparent and efficient learning and cognitive computation.”
As director of the Bio-Inspired Architecture and Systems Lab (BIASLab), Imani has a team of 35 members, including 15 Ph.D. students working on multiple interdisciplinary research projects to bridge the gap between neuroscience, machine learning and embedded systems. For this DARPA project, Imani and his team are developing a comprehensive and cross-layer brain-inspired model that bridges human cognition and the underlying neurobiological mechanisms. The goal is to develop a framework for efficient and highly scalable learning systems that can analyze large-scale data at the source.
“In today’s systems, we face increasing needs for efficient and real-time processing for diverse cognitive tasks using a vast volume of generated data,” explains Imani. Although several machine learning algorithms have been developed to enable such intelligence, these models are not explainable and require significant computational power and storage. “As a result, many devices stream most data to the cloud for analysis, which leads to scalability, security and privacy concerns.”
To address this, Imani and his team are designing a novel system with human-like learning and memorization capabilities. “The developed algorithms rely on learning and compositionality, where the compositional structure enables recognizing complex concepts by symbolically combining the primitives,” he says. “This project is a step towards a viable solution to mitigate brain-like intelligence to design efficient and reliable learning systems.”
Imani has a track record of several successful technological transfers to industry and government. In fact, in a June newsletter, Semiconductor Research Corp. highlighted Imani’s contributions to multiple top semiconductor industries, including Intel, IBM, NXP and Qualcomm. He was also selected as a DARPA Riser in 2022 — a program that recognizes up-and-coming standouts in their fields.
Imani hopes his DARPA YFA project helps revolutionize numerous Internet of Things applications, allowing low power devices to not only learn more efficiently but also adapt to new information without requiring time-consuming training. “We hope to enable future computing systems that operate quickly and efficiently and can easily adapt to new information while preserving the accuracy and error bounds.”
— Shani Murray