The videos of the talks are available on the Videolectures website,
If you couldn't make it to COLT 2014 or if you want see the talks again, now is your chance!Tuesday, July 29, 2014
Date/time: June 11th from 15:30 to 18:00 (coffee break from 16:30 to 17:00), and June 12th from 14:00 to 16:30 (coffee break from 15:00 to 15:30).
Location: Centre de Recerca Matemàtica, main auditorium [note: different location from COLT!]
Title: Analyzing Deep Neural Networks for Learning
Abstract: There is a big buzz around deep neural networks, with impressive numerical results, but nearly no mathematical back up explaining these performances. Available mathematical approaches to circumvent the curse of dimensionality, typically do not apply to classification of complex high-dimensional data such as signals or images, or to regressions of non-local physical functionals. Beyond applications, deep networks may thus be an opportunity to develop new mathematics for high dimensional problems.
The two lectures will introduce some mathematical tools to analyze these networks, and describe numerical results, while encouraging discussions on open questions. The following topics will be covered:
- Curse of dimensionality and approximation theory
- One hidden layer neural network
- Reduction of dimension and contractions
- Multiscale wavelets and stable invariants over Lie groups
- Deep convolution networks and scattering operators
- Regression of physical functionals and N-body problems
- Image and audio classification
- Unsupervised and supervised deep learningWednesday, June 4, 2014
Each year two new members of the COLT steering committee (formally "Board of Directors of the Association for Computational Learning") are elected at the COLT conference.
To nominate candidates for this election send candidate names to firstname.lastname@example.org before 4 June 2014 (self nominations are fine). Please verify that the candidates are willing to serve on the steering committee if elected.
Chair of the COLT steering committeeThursday, May 22, 2014
Monday, April 21, 2014
Monday, March 10, 2014
COLT 2014 will include a session devoted to the presentation of open problems. A description of these problems will also appear in the COLT proceedings. The deadline for submissions is April 10.
The write-up of an open problem should include:
-- a clearly defined theoretical problem.
-- the motivation for studying the problem, with an argument why it is important, interesting, and not too easy.
-- the current state of this problem (including any known partial or conjectured solutions and relevant references).
As last year, we encourage submissions of problems not conventionally in the scope of COLT, as long as there is a convincing reason to include it in the conference. You should be able to clearly express the problem in a short presentation. A monetary reward for solving an open problem is encouraged but not required.
If you would like to submit a problem, please email it to email@example.com by April 10, 2014. Submissions should be at most 3 pages long and should be in the COLT'14 JMLR format (pdf or ps). Submissions should be non-anonymous.
Examples of open problems from past years:
2012: http://jmlr.org/proceedings/papers/v23/ (scroll down)
2011: http://jmlr.csail.mit.edu/proceedings/papers/v19/ (scroll down)
2009: http://www.cs.mcgill.ca/~colt2009/proceedings.html (scroll down)Sunday, March 2, 2014
Detailed instructions for submission have been posted.
We are also happy to announce that COLT 2014 will take place within a larger research program on the Mathematics of Machine Learning organized at the Centre de Recerca Matemática in Barcelona.Tuesday, January 14, 2014
The 27th Annual Conference on Learning Theory (COLT 2014) will take place in Barcelona, Spain, on June 13-15, 2014.
We invite submissions of papers addressing theoretical aspects of machine learning and related topics. Submissions by authors who are new to COLT are encouraged. We strongly support a broad definition of learning theory, including, but not limited to:
• Design and analysis of learning algorithms and their generalization ability
• Computational complexity of learning
• Optimization procedures for learning
• Unsupervised, semi-supervised learning, and clustering
• Online learning
• Interactive learning
• Kernel Methods
• High dimensional and non-parametric empirical inference, including sparsity methods
• Planning and control, including reinforcement learning
• Learning with additional constraints: E.g. privacy, time or memory budget, communication
• Learning in other settings: E.g. social, economic, and game-theoretic
• Analysis of learning in related fields: natural language processing, neuroscience, bioinformatics, privacy and security, machine vision, data mining, information retrieval.
We are also interested in papers that include viewpoints that are new to the COLT community. We welcome experimental and algorithmic papers provided they are relevant to the focus of the conference by elucidating theoretical results. Also, while the primary focus of the conference is theoretical, papers can be strengthened by the inclusion of relevant experimental results.
COLT will award both best paper and best student paper awards. Best student papers must be authored or coauthored by a student. Authors must indicate (using a footnote on the first page of the paper) at submission time if they wish their paper to be eligible for a student award. This does not preclude the paper to be eligible for the best paper award. The program committee may decline to make these awards, or may split them among several papers.
Papers that have previously appeared in journals or at other conferences, or that are being submitted to other conferences, are not appropriate for COLT. Papers that include work that has already been submitted for journal publication may be submitted to COLT, as long as the papers have not been accepted for publication by the COLT submission deadline (conditionally or otherwise) and that the paper is not expected to be published before the COLT conference (June 2014).
Accepted papers will be published electronically, in the JMLR Workshop and Conference Proceedings series.
As in the previous years, we will have a rebuttal phase during the review process. Initial reviews will be sent to authors before final decisions have been made. Authors will have the opportunity to provide a short response on the PC's initial evaluation. Final acceptance/rejection decision will be made a week later.
Open Problems Session:
We also invite submission of open problems. A separate call for open problems will be available at the conference website.
Submissions are limited to 12 JMLR-formatted pages, plus additional pages for references and appendices. All details, proofs and derivations required to substantiate the results must be included in the submission, possibly in the appendices. However, the contribution, novelty and significance of submissions will be judged primarily based on the main text (without appendices), and so enough details, including proof details, must be provided in the main text to convince the reviewers of the submissions' merits.
Formatting and submission instructions can be found here.
• Paper submission deadline: February 7th, 2014, 11:00 PM EST
• Author response period: April 5-10, 2014
• Author notification: April 19, 2014
• Conference: June 13-15, 2014Wednesday, September 25, 2013
The 27th Annual Conference on Learning Theory (COLT 2014) will take place in Barcelona, Spain, on June 13-15, 2014.Friday, June 21, 2013