Research Areas
Machine Learning, Robustness, Optimization, and their Applications. Current research focus includes:
Robustness: Building theoretical foundations for computationally efficient adversarial defenses and attacks. Developing practical, large-scaled algorithms for real-world AI security problems.
Optimization: Developing new paradigms toward global optimality of non-convex optimization in polynomial time. Designing algorithms and understanding landscape (e.g., duality gap) of deep neural network, GAN, matrix factorization.
Sample Efficiency: Designing principled, practical and scalable algorithms for big data problems with near-optimal sample complexity. These include models of matrix completion and sensing, robust PCA, margin-based active learning, property testing, phase retrieval.
Applications: Applications of machine learning models in image and video processing, medical data.