Machine Learning

This lecture discusses online learning and how we might quantify performance of an online learning algorithm using the mistake bound approach.

Lecture slides

• Online Learning (Mistake bound learning)

• Videos:

  • Lecture 1 introduces the high level idea
  • Lecture 2 defines the mistake bound model and the Halving algorithm

  • Lecture 3 discusses the representations and learnability

  • Older videos:
    • [spring 2023, part 1], [spring 2023, part 2], [spring 2023, part 3]
    • [fall 2018, part 1], [fall 2018, part 2],
    • [fall 2017]

Links and Resources

• Chapter 7.5 of Tom Mitchell’s book discusses the Halving algorithm

• Avrim Blum’s notes on the mistake bound model

Online learning books

Both these books are available online and offer detailed exposition about online learning.

• Shalev-Shwartz, Shai. “Online learning and online convex optimization.” Foundations and trends in Machine Learning 4, no. 2 (2011): 107-194.

• Hazan, Elad. “[Introduction to Online Convex Optimization.](https://arxiv.org/abs/1909.05207” Foundations and Trends in Optimization 2, no. 3-4 (2016): 157-325.