A field concerned with analysis of learning algorithms, finding:

  • Generalization guarantees
  • Regret bounds Shows that a learning algorithm under certain conditions is guaranteed to achieve some level of performance.