The MLOPT research group is building the mathematical and computational foundations of machine learning and optimization theory. Our work aims to answer fundamental questions about

  • the performance limits of machine learning systems
  • tradeoffs between data size, computational complexity, and statistical accuracy
  • the mathematical characterizations of functions and representations learned from data
  • computational and communication complexity in distributed learning and optimization
  • privacy and fairness in machine learning and data science
  • the interpretability, scalability, and safety of algorithms