COLT 2027 — 40th Annual Conference on Learning Theory

COLT 2027, the 40th Annual Conference on Learning Theory, runs 28 June – 2 July 2027 in Tokyo, Japan. The Association for Computational Learning, which has organised COLT annually since 1988, describes it as "the leading conference on the theory of machine learning and artificial intelligence" and runs "a highly selective and rigorous review process". Submissions cover theoretical aspects of machine learning "broadly defined as a subject at the intersection of computer science, statistics and applied mathematics": design and analysis of learning algorithms, statistical and computational complexity of learning, theory of artificial neural networks including deep learning, reinforcement learning, learning with system constraints such as privacy, fairness, memory and communication, and quantum learning theory. Programme chairs, submission deadlines and the Tokyo venue were all still listed as to be determined as of September 2026.

Date

June 28 – July 2, 2027

Location

Tokyo, Japan

Format

In-Person

Event Type

Conference

Status

Upcoming

Pricing

Paid

Organizer

Association for Computational Learning

Added

Sep 7, 2026

Updated

Sep 7, 2026

• ✓Machine learning theorists
• ✓AI research scientists
• ✓Statisticians
• ✓Research directors at industrial AI labs
• ✓Quantum computing researchers
• ✓PhD students in learning theory

COLT is where the guarantees behind production machine learning are established years before they surface as vendor claims. For a CTO or head of AI, the value is early sight of what is provably achievable under the constraints enterprises actually operate in — differential privacy, fairness, bounded memory and communication, and the sample complexity of reinforcement learning from human feedback. It is also the densest recruiting pool anywhere for theory-trained researchers, and the 2027 edition puts that pool in Tokyo, within short-haul reach of Asia-Pacific engineering and research teams.

• ✓Topic area: design and analysis of learning algorithms
• ✓Topic area: statistical and computational complexity of learning
• ✓Topic area: theory of artificial neural networks, including deep learning
• ✓Topic area: reinforcement learning
• ✓Topic area: learning with system constraints — privacy, fairness, memory, communication
• ✓Topic area: quantum learning theory
• ✓Double-blind review with a rebuttal phase; papers limited to 12 PMLR-formatted pages excluding references and appendices

• ★The 40th edition — COLT has run annually since 1988, making it one of the longest-running venues in machine learning.
• ★The Association for Computational Learning calls it "the leading conference on the theory of machine learning and artificial intelligence" and runs "a highly selective and rigorous review process".
• ★Named topic areas include quantum learning theory and learning under privacy, fairness, memory and communication constraints — subjects most applied machine learning venues do not carry at all.
• ★The organisers "strongly support an inclusive view of learning theory" and explicitly invite work giving a "theoretical explanation of empirical phenomena in learning", rather than bounds alone.

Registration is not open and no pass prices have been published for the 2027 edition as of September 2026. The official COLT 2027 site lists every call-for-papers milestone — submission, reviews, author response, discussion period and notification — plus the programme chairs and the venue, as still to be determined.

When is COLT 2027?

COLT 2027 takes place June 28 – July 2, 2027 in Tokyo, Japan.

Where is COLT 2027 held?

COLT 2027 is held in Tokyo, Japan. The event format is in-person.

How much does COLT 2027 cost?

COLT 2027 is paid. Registration is not open and no pass prices have been published for the 2027 edition as of September 2026. The official COLT 2027 site lists every call-for-papers milestone — submission, reviews, author response, discussion period and notification — plus the programme chairs and the venue, as still to be determined.

Who should attend COLT 2027?

COLT 2027 is aimed at Machine learning theorists, AI research scientists, Statisticians, Research directors at industrial AI labs, Quantum computing researchers, PhD students in learning theory.

What topics does COLT 2027 cover?

COLT 2027 covers Learning theory, Statistical learning, Computational complexity of learning, Deep learning theory, Reinforcement learning theory, Privacy and fairness constraints, Quantum learning theory, Online learning.

This page was written from 4 sources, 1 on domains other than learningtheory.org.

1. 1.learningtheory.org — colt2027vendor
2. 2.learningtheory.org — learningtheory.orgvendor
3. 3.learningtheory.org — cfpvendor
4. 4.opencurious.com — ai conference deadlines