FBBVA-Ayudas-Equipos-2016-Pedro-Larranaga

‘Multi-view clustering with Bayesian networks’

Grants for research teams

Big Data

2016

The main objective of this project is the investigation, development and innovation on multi-view clustering with Bayesian networks and their application with streaming data on a big data scale. Specific objectives can be categorized into methodological, technological, and applications.

DIRECTOR

Pedro Larrañaga Mugica, professor in Artificial Intelligence at the Technical Universidad Politécnica de Madrid (UPM)

 

RESEARCH TEAM

Concha Bielza Lozoya; Juan Antonio Fernández del Pozo de Salamanca; Gherardo Varando; Sergio Luengo Sánchez; Luis Rodríguez Luján; Marco Benjumeda Barquita and Irene Córdoba Sánchez, Technical University of Madrid; José Manuel Peña Paloma, Linköping University; Joao Manuel Portadela da Gama, University of Porto.

COLLABORATING INSTITUTIONS

Universidad Politécnica de Madrid (UPM)

 

DESCRIPTION

The main objective of this project is the investigation, development and innovation on multi-view clustering with Bayesian networks and their application with streaming data on a big data scale.

Specific objectives can be categorized (see below) into methodological (1 and 2), technological (3), and applications (4 and 5).

A list of these specific objectives and sub-objectives follows:

1. New latent tree models learning algorithms for discrete, continuous and hybrid data.
     1.1 New latent tree models learning algorithms for discrete and continuous data.
     1.2 New latent tree models learning algorithms for hybrid (discrete and continuous) data.

2. Multi-view clustering of streaming data.
     2.1 Latent tree models for the stream paradigm.
     2.2 Adaptive models for multi-view clustering in the stream paradigm.

3. Develop a reusable and accessible framework that would be useful when dealing with big data challenges.
     3.1 Design and implement algorithms for parallel processing of distributed and centralized streaming data.
     3.2 Create this machine learning framework on top of well-known big data technologies for  easier reusability and better performance.
     3.3 Develop a visual tool for real-time monitoring of the generated adaptive models.
     3.4 Distribute this framework on a public repository.
     3.5 Investigate the combination of NoSQL databases (i.e Apache Cassandra) with the developed framework for storing streaming data.

4. Applications of multi-view clustering in neuroscience.
     4.1 Multi-view clustering of Parkinson’s patients.

5. Applications of multi-view clustering in the industry.
     5.1 Multi-view clustering of sensor network´s data.