Big Data : Applications and Algorithms

Date: October 12th, 2012

Organizers

Dr José Luis Fernandez-Marquez (Post-Doc)

Prof. Giovanna Di Marzo Serugendo

University of Geneva, Facultés des SES

Speakers
Dr. Francisco J. Martin (CEO BigML, Oregon, USA)
Ivan de Prado (CEO DataSalt, Barcelona, Spain)
Pere Ferrera (CTO DataSalt, Barcelona, Spain)

Description
Dr. Francisco J. Martin (CEO BigML, Oregon, USA)
Title: BigML - Machine Learning Made Easy (3h)

In the "Big Data" era, rapidly and easily getting insights from your data or creating data-driven applications does not have to be painful. BigML is working to make machine learning extremely easy to use and seamless to integrate. We will show you how BigML can help business managers, application developers, and data scientists in data-rich domains build their own predictive models in a matter of minutes.

Ivan de Prado (CEO DataSalt, Barcelona, Spain)
Pere Ferrera (CTO DataSalt, Barcelona, Spain)
Title: Tuple MapReduce: Beyond classic MapReduce (3h)

During the last years,  the amount of information handled within different fields (i.e. webs, sensor networks, logs, or social networks) has increased dramatically.

Well established approaches, such as programming languages, centralised frameworks, or relational databases, do not cope well with current companies requirements arising from needs for higher-levels of scalability, adaptability, and fault-tolerance. These requirements are currently demanded by many companies and organisations that need to extract meaningful information from huge volumes of data and multiple sources. Even though many new technologies have been recently  proposed for processing huge amounts of data, there is still room for new technologies that combine efficiency and easiness in solving real-world problems.

In this seminar we introduce Tuple MapReduce, a new foundational model extending MapReduce with the notion of tuples.  Tuple MapReduce allows to bridge the gap between the low-level constructs provided by MapReduce and higher-level needs required by programmers, such as compound records, sorting or joins.

We present also Pangool, an open-source framework implementing Tuple MapReduce. Pangool eases the design and implementation of applications based on MapReduce and increases their flexibility, still maintaining Hadoop's performance.