Evolutionary Computation for Big Data and Big Learning Workshop

Important dates:

• Paper submission deadline: April 3rd, 2015
• Author notification: April 20, 2015
• Camera-ready version: May 4, 2015
• Registration to Big Learning activity: TBD
• Big Learning activity period: TBD


• Workshop date: July 11-15, 2015
• Workshop Location: TBD

Workshop outline

We live in a time of unprecedented access to cheap and vast amounts of computational resources, which is driving a big leap forward in the fields of machine learning and data mining. Big learning techniques allow us to tackle datasets of scales (be they instances, attributes, classes, etc.) that were unimaginable some years ago. We can also use these vast computational resources to obtain more predictive features or to better understand our machine learning methods, by performing large scale evaluations, parameter sweeps, etc. We refer to the use of massive on-demand computation (cloud or GPUs) for machine learning as Big Learning. Evolutionary Machine Learning techniques remain good candidates for Big Learning given their flexible representation, parallelism, and implicit variable selection techniques.

The overall objective of this workshop is to assess the state of the art in evolutionary computation methods for Big Learning, with special focus on feature extraction and large datasets. To achieve this aim we make a call for speakers to present their big data/big learning methods at the workshop.

Call for papers

The topics covered by this workshop, within the scope of evolutionary computation, are (but are not limited to):

• large-scale feature learning
• systems for large scale machine learning
• large ensemble learning in parallel
• parameter sweeps in parallel
• machine learning for high dimensional data
• learning many-class classification problems (extreme class problems)
• parallel cross-validation at scale
• adaptive learning algorithms and online parameter tuning
• Machine Learning for Big Data
• GPGPUs for Machine Learning

Authors are highly encouraged to demonstrate their approaches with one (or more) of the following datasets:

• Classification: Higgs binary classification dataset
• Regression: Million song dataset year prediction challenge
• Feature learning: Kaggle malware classification challenge. Note that the authors do not need to participate in the competition, just to use the data to demonstrate their approach.

Please submit your contributions by email to iarnaldo@mit.edu. Your contribution should be formatted according to the ACM guidelines, in PDF format, and not exceed 8 pages. Unlike in the main GECCO conference, papers do not need to be submitted in anonymized form.

Call for Big Learning activity participants

TBD

Organizers

Ignacio Arnaldo: iarnaldo@mit.edu
Kalyan Veeramachaneni: kalyan@csail.mit.edu
Una-May O'Reilly: unamay@csail.mit.edu
Jaume Bacardit: jaume.bacardit@newcastle.ac.uk