- This event has passed.
Physics and Astronomy Colloquium: Sustainable, Reliable, and Secure AI
February 27 @ 3:15 pm – 4:30 pm
The latest installment of the Colloquium Series from the Department of Physics and Astronomy is scheduled to take place on February 27 at 3:15 PM in 140 Bennett Hall.
The event – titled “Sustainable, Reliable, and Secure AI” – will be presented by Ayesha Siddique, assistant professor in the Department of Computer and Electrical Engineering at the 91±¬ÁÏ.
From the abstract:
Approximate computing has been widely explored as a means to improve the energy efficiency of deep neural networks (DNNs) for edge AI hardware. However, both accurate and approximate DNNs remain inherently vulnerable to faults and adversarial attacks, and the reliability and adversarial robustness of approximate DNNs remain largely unexplored. This gap presents an opportunity to rethink how we design sustainable and dependable AI hardware. In this seminar, I will discuss our recent advances in post-fabrication fault mitigation methods and the principled selection of hardware approximation techniques to preserve adversarial robustness under attack. I will highlight how we leverage emerging computing paradigms, such as explainable artificial intelligence, neural architecture search, moving target defense, and neuromorphic computing, to ensure reliability, security, and energy efficiency. This seminar outlines a path toward AI hardware that is not only high-performing and energy-aware but also resilient against both hardware faults and security threats.
This event is free and open to the public.
