[Liste-proml] ICONIP'12 Special Session on Co-clustering of Large and High Dimensional Data : call for papers

Nicoleta.Rogovschi at mi.parisdescartes.fr Nicoleta.Rogovschi at mi.parisdescartes.fr
Mer 2 Mai 09:04:33 CEST 2012

Dear Colleagues,
2012 International Conference on Neural Information Processing (ICONIP  
2012) will be held in Doha, Qatar. It is an annual event, organized  
since 1994 by the Asia Pacific Neural Network Assembly (APNNA).

Special Session : Co-clustering of Large and High Dimensional Data

Younès Bennani, Full Professor, Paris 13 University.
Nistor Grozavu, Associate Professor, Paris 13 University.
Mohamed Nadif, Full Professor, Paris 5 University.
Nicoleta Rogovschi, Associate Professor, Paris 5 University.

Nistor Grozavu, Nicoleta Rogovschi
Email: Nistor.Grozavu at lipn.univ-paris13.fr,  
nicoleta.rogovschi at parisdescartes.fr

  One of the strongest problems afflicting current machine learning  
techniques is dataset dimensionality. During the recent years,  
companies, collaborations, and organizations are increasingly faced  
with the need to analyze of large and growing collections of data. Big  
data requires technologies to efficiently process large quantities of  
data within tolerable elapsed times. The used technologies include  
massively parallel processing datasets, multi-sourcing, cloud  
computing platforms, distributed databases...

Beyond the purely quantitative, these data are presented in such a way  
that they are hardly supported by traditional machine learning  
methods: data is not organized in the form of tables and their  
structures can vary; is produced in real time; arrive worldwide in  
continuous streams; is heterogeneous and located on different sources  
(mobile phones, sensors, connected TVs, tablets, desktop PCs, laptops,  
objects, machines).

In particular, the high dimensionality of data is a highly critical  
factor for the clustering task. Another important challenge in  
clustering is the dimensionality reduction which deals with the  
transformation of a high dimensional data set into a low dimensional  
space, while retaining most of the useful structure in the original  
data; retaining only relevant features and observations. Different  
clusters might be found in different subspaces, so a global filtering  
of attributes is not sufficient. Although the majority of clustering  
procedures aim to construct an optimal partition of observations or,  
there are other methods, known as co-clustering in document  
clustering, biclustering in bioinformatics context or latent block  
models, which consider the two sets simultaneously and organize the  
data into homogeneous blocks. Compared to conventional clustering  
methods, they have proved their effectiveness in discovering  
structures from matrices of large high dimensional datasets.

  All these challenges can be treated together in order to establish a  
scalable, affordable, and flexible large-scale analytics infrastructure.

  The aim of this Special Session is to provide a forum for  
discussions of recent advances and future directions in the  
Co-clustering of Big Data and its topic areas include, but are not  
limited to:

  - subspace clustering
  - high dimensionality reduction
  - latent block models
  - model selection
  - multi-view co-clustering
  - collaborative co-clustering
  - distributed co-clustering
  - multi source co-clustering
  - consensus co-clustering
  - Nonnegative Matrix Factorization
  - incremental co-clustering
  - visualization
  - high sparse data co-clustering

Yours sincerely,
Nicoleta Rogovschi.

This message was sent using IMP, the Internet Messaging Program.

Plus d'informations sur la liste de diffusion Liste-proml