With the burgeoning use of artificial intelligence and machine learning (AI/ML) models, such as Large Language Models (LLMs), Vision Language Models (VLMs) and generative AI, in an assortment of applications, there is a need to learn and deploy models, that align with human values, in an ever-changing world rapidly and reliably. In this talk, I will cover my group’s recent research focusing on robustness, efficiency, and fairness in AI/ML models, vital in fostering an era of Trustworthy AI that society can rely on. Our efforts on fortifying models against (adversarial/distribution) shifts and spurious features, enhancing model, data, and learning efficiency, and ensuring long-term fairness under distribution shifts will be discussed
Dr. Furong Huang is an Assistant Professor of the Department of Computer Science at the University of Maryland. Her research focuses on machine learning, high-dimensional statistics, non-convex optimization, spectral methods, reinforcement learning and deep learning theory. The thesis of Dr. Huang’s research is to understand the foundations of deep learning under distribution shifts, adversarial perturbations and non-iid data. Through the investigation of principled methods that address the modern challenges in applying ML to real-world application, Dr. Huang’s group expands the scope of deep learning model design for learning in constrained edge clients and on graph data.