Current Work

  • Responsible AI (safety, robustness, etc.) – see this ICML tutorial and this talk for more details.
  • Uncertainty quantification for AI systems and its applications in safety-critical domains (e.g., autonomous driving, healthcare, etc.).
  • (Neural) data compression.
  • Mathematical foundations of deep learning.
  • Submodular optimization and its applications in machine learning – see this ICML tutorial (multiple parts) for more details.
  • Distributed / federated learning.
  • Modern coding theory.
  • (Large-scale) probabilistic models.
  • Graph matching: theory and algorithms.