COLT 2026 — Conference on Learning Theory
📅 Monday, 29 June 2026 → Friday, 3 July 2026
COLT 2026 — the 39th Annual Conference on Learning Theory, San Diego, 29 June–3 July.
COLT 2026, the 39th Annual Conference on Learning Theory, takes place 29 June–3 July 2026 at the Bahia Resort Hotel in San Diego, organised by the Association for Computational Learning. It is the leading venue for the mathematical foundations of machine learning.
What you gain
COLT is where the rigorous, proof-driven understanding of why and when learning works is presented — generalisation, optimisation, online learning, and the theory behind today's large models. Programme chairs are Steve Hanneke (Purdue) and Tor Lattimore (Google DeepMind), with an Open Problems session that seeds future research directions. For theorists and theoretically-minded ML researchers it is the field's intellectual centre of gravity.
Where it fits
Amid an empirically-driven AI boom, COLT supplies the foundations that explain and constrain it. Attending puts you at the source of the ideas — sample complexity, optimisation guarantees, learning under uncertainty — that quietly shape what the rest of the field builds.
