The Extreme Classification workshop will take place July 10th 2015 in conjunction with the International Machine Learning Conference (ICML) in Lille, France.
Extreme Classification (i.e. multi-class or multi-label classification in the presence of an enormous number of labels) has recently attracted increased attention in the machine learning community and beyond (computer vision, NLP, bioinformatics). Indeed, several applications such as recommendation, ranking, etc. can be reduced to extreme classification problems. This new perspective allows to leverage the recent advances achieved in reducing prediction time and making learning feasible in extreme classification problems. However, the issues posed by extreme classification go beyond computational complexity. In particular it is one of the rare situations where data scarcity remains a question despite the vastness of available data. Also, the statistical dependence/correlation of the labels poses challenges and opportunities for learning approaches. All the properties of extreme classification problems (e.g. data and feature distribution) and their specificities across different fields are still not well understood. Moreover, questions related to valid evaluation measures in this setting have rarely been discussed despite being widely considered as critical. In particular, we are interested in questions related to:
The Extreme Classification workshop has 4 confirmed invited speakers
More speakers will be added upon confirmation
We welcome submissions on novel research as well as extended abstracts on relevant work recently published in venues of related fields (computer vision, computational biology) or in journals. In case of previously published work, the original paper should be clearly cited. We also welcome promising work-in-progress contributions and position papers. All accepted papers will be presented as posters, and some will be selected for oral presentation.
Submissions should follow the ICML format and are encouraged to be between four and eight pages long. Reviewing will be single-blind (submissions need not be anonymous). Accepted papers will be posted online, but there will be no proceedings. Please submit papers in PDF format here submission site. For any question, send emails to firstname.lastname@example.org