Deliang Fan has nothing less than bold expectations for his endeavors to expand the capabilities of computing systems and technologies. His far-reaching aspirations involve designing advanced computational hardware to help innovate in some of the most technologically complex areas of engineering and science.
He especially wants to make significant advances in high-performance, energy-efficient computing for big data processing and make big steps in developing more vigorous artificial intelligence, or AI, computing.
Fan, an electrical and computer engineer and assistant professor in the Ira A. Fulton Schools of Engineering at Arizona State University, says his ultimate goal is to “design, implement and conduct the experimentation to validate the performance of a new hybrid in-memory computing system.”
In-memory computing is a way of running computer calculations entirely in computer memory. For example, random-access memory, or RAM, is short-term memory where data is stored as needed by a computer’s processor.
“The key concept is to leverage the memory device or circuit properties to implement logic functions within a memory array to directly process stored data,” Fan says, “and to do this without moving data to a separate central logic unit, thus minimizing energy- dominating data communication. ”
The idea, he says, is to “optimize energy efficiency, inference accuracy, spatiotemporal dynamics, robustness and on-device learning, which will greatly advance AI-based big data processing fields, such as computer vision, autonomous driving and robotics.”
Based on his progress so far, the National Science Foundation, or NSF, is convinced Fan is prepared to take on these formidable challenges. He was recently awarded a 2022 NSF CAREER Award that will provide $ 500,000 over five years to fund his research.
The CAREER Award program supports university faculty members early in their careers who are deemed to have the potential to excel in both research and education, to serve as role models in their academic departments and to lead advances in fields in which progress will serve national interests and priorities.
Fan’s Efficient, Secure, and Intelligent Computing Laboratory team is attempting to build “a revolutionary and ultra-efficient ‘AI-in-memory’ computing system that could execute AI computation directly within the memory where the data is stored, without the massive data movement , ”Says Fan, who teaches in the School of Electrical, Computer and Energy Engineering, one of the seven Fulton Schools.
If the project is successful, it will provide a new kind of in-memory computing architecture that will operate faster and be a hundred times more energy efficient than current state-of-the-art central processing units or graphics processing units.
“This tiny but powerful computing system will bring computing solutions to the existing smart Internet of Things technologies, robots, self-driving cars, smart-connected health technologies, on-device learning capability and many other things,” Fans says. “It will provide a much cheaper, faster and low-power AI computing platform.”
The project will include extensive educational components. Fan plans to recruit both undergraduate and graduate students to participate in conducting research for the project.
More educational efforts will be made through multiple K – 12 STEM outreach programs designed to help younger students learn the basics about computer engineering and AI fields.
In addition, a new course in the growing area of in-memory computing is being developed, with plans to offer the course in the School of Electrical, Computer and Energy Engineering.