Connectionists: Call For Contributions: NIPS 2011 workshop on Challenges in Learning Hierarchical Models: Transfer Learning and Optimization

[Ruslan Salakhutdinov]

CALL FOR CONTRIBUTIONS IN CHALLENGES

NIPS 2011 workshop on Challenges in Learning Hierarchical Models: Transfer 
Learning and Optimization

Melia Sierra Nevada & Melia Sol y Nieve, Sierra Nevada, Spain

https://sites.google.com/site/nips2011workshop/home


Important Dates:
----------------

Deadline for submissions:          October 17, 2011
Notification of acceptance:        October 28, 2011


Overview:
----------------

As part of the 2011 workshop on Challenges in Learning Hierarchical 
Models, we also organize two challenges. The winners of each challenge 
will be awarded an oral presentation of their method at the NIPS workshop.

*Optimization challenge:* Participants are asked to submit code to train 
neural networks with various depths on a large dataset. At evaluation 
time, we will vary slightly parameters of the networks and benchmark the 
training time. Participants will have a lot of freedom, such as the use of 
pretraining, initialization, programming languages and so on. More details 
regarding this challenge are available at

https://sites.google.com/site/nips2011workshop/optimization-challenges

*Transfer learning challenge:* Participants are given a large dataset 
(CIFAR-100), which they can use to train their models on, as well as a 
related task with very few labelled examples. At evaluation time, models 
will be evaluated on this related task. The examples in the related task 
have been recently collected as part of this challenge and have never 
appeared in the literature.

https://sites.google.com/site/nips2011workshop/transfer-learning-challenge


Submission Instructions:
----------------

We solicit participants to test their methods on two challenges:
- Transfer Learning Challenge
- Optimization Challenge

The aim is to facilitate discussion among participants through a common 
task using a strict protocol that makes results comparable. Results should 
be submitted via email to: transflearn.optim.wnips2011 at gmail.com no later 
than 23:59 EST, Monday, October 17, 2011.

The paper and challenge tracks are independent from each other. This means 
that the contributors can choose to submit a paper to the workshop without 
contesting in challenges or contest in one of the challenges without 
submitting. Multiple submissions to different tracks (paper and challenge) 
are also encouraged.

Organizers
----------
Quoc V. Le, Computer Science Department, Stanford University
Marc'Aurelio Ranzato, Google Inc
Ruslan Salakhutdinov, Department of Statistics, University of Toronto
Andrew Ng, Computer Science Department, Stanford University
Josh Tenenbaum, Department of Brain and Cognitive Sciences, MIT