Connectionists: Workshop on data representation discovery

[Isabelle Guyon]

                        IJCNN07 Workshop on Data Representation Discovery
                                        Thursday, August 16, 2007
                                                Orlando, Florida

Paper submission deadline: January 31st 2007
URL: http://clopinet.com/isabelle/Projects/agnostic/

With the proper data representation, learning becomes almost trivial! 
Topics of interest include, but are not limited to:

- space embedding, space dimensionality reduction, latent variable methods
- preprocessing (noise modeling, spectral transformations, etc.)
- feature extraction (including feature construction and  feature 
selection)
- dynamic representations (selective attention; top down expectancy with 
bottom-up evidence resolution; dynamic feature selection)
- biologically/psychologically inspired data representations
- data representations for image, speech, and video processing
- data representations for molecules, including applications in drug 
discovery and bioinformatics
- data representations for text processing, including language translation
- supervised, unsupervised, semi-supervised methods of learning data 
representations
- learning representations with neural networks
- learning representations with evolutionary computing
- learning kernels (in kernel methods), learning similarity measures, 
learning distance metrics
- transformation invariant representations, robust representations
- multi-level data representations
- multi-objective data representations
- theoretical and empirical assessement of data representations

The emphasis will be more on generic principled methods rather than on 
ad-hoc solutions.
The results of the "Agnostic Learning vs. Prior Knowledge" challenge 
will be discussed at the workshop. The challenge is open until March 
1st, 2007: http://www.agnostic.inf.ethz.ch/. Win one of seven prizes!