ECE Faculty Awarded Funding for AI Hardware

A group of ECE faculty members, led by Professor Sahil Shah, has been awarded $300,000 in funding over three years from SRC, Inc., an independent, not-for-profit research and development company. Don Yeung and Dinesh Manocha are Co-PIs on the project. Their project, “Hardware-Algorithm co-design for on-chip learning at the edge” was selected as part of the SRC 2023 Artificial Intelligence Hardware Program

Data acquired by edge devices requires analysis using machine learning algorithms. Shah, Yeung, and Manocha argue that devices used in areas such as remote monitoring, sensing, continuous vital signal tracking and autonomous robots have limited computational resources, including small battery capacities and constrained memory for storing and processing gradients. Through this project, they will focus on creating more energy-efficient circuits and architectures using analog non-volatile devices for computing neural networks. In addition, they will explore circuits that can enable on-chip learning capabilities to be applied to numerous commercial applications.

The project will address the following research direction:

Modeling non-linear mixed-signal circuits for computing neural networks and exploring efficient architectures using the models for running tiny ML benchmarks.
Modeling variations observed in mixed-signal circuits and studying their effects on accuracy of neural networks.
Exploring circuits that reduce the overall power consumption and area for in-memory computing architecture.

Ultimately, advancements to on-chip learning at the edge through solid-state circuits and architecture can impact multiple industries, including manufacturing, healthcare, autonomous vehicles, energy management and security systems. Creating tailored solutions can improve efficiency, enhance privacy, security and real-time decision making in industrial IoT, autonomous vehicles, energy management and security systems.

In alignment with their commitment to fostering educational growth and practical experience, the team, led by Professors Sahil Shah, Don Yeung, and Dinesh Manocha, is dedicated to integrating both graduate and undergraduate students into the core of their research. This initiative not only provides these students with a unique opportunity to contribute to groundbreaking work in cutting-edge technology but also serves as an invaluable mentoring experience. Under the guidance of experienced faculty, these students will gain hands-on expertise in specialized areas such as mixed-signal circuits, the development of energy-efficient architectures, and the intricacies of hardware-aware neural network algorithms. This approach not only enriches their academic journey but also prepares them for future challenges and innovations in the fields of artificial intelligence and hardware design.

Published February 2, 2024