Seminar URL : http://community.liglab.fr/students/social-events/winter-seminar/

 Submission date : 2 April 2011

 Seminar date : 4 May 2011

 Contact us : phd-synergy@imag.fr

 See also : Life of the PhD students

 Programme

— 1:30 PM Accueil - coffee

— 2:00 PM Ouverture Rémi Barraquand & Emeric Fontaine

Research in the industry

— 2:15 PM Eustache Diemert, ex-Yahoo ! Search, now Lead Scientist at BestOfMedia

• Title : Industrial Applications of Machine Learning : Web & Data Sciences
• Summary : Machine Learning is nowadays more and more common in the industry, especially in domains where data is central to business efficiency. There is plenty of examples of application of Machine Learning for Web applications : Search Engines, Spam Detection software, Social Networks Mining, User Tracking & Modeling, Topic Modeling etc. all of which have seen much interest from both academic and industrial researchers since the beginning of the century. I will present a few Web applications with insights on the models, techniques, algorithms used in each setting and try to highlight the challenges faced by scientists working on these applications.

Ph.d Student talks

— 3:00 PM Ali Mustafa Qamar, ATER at UJF, Grenoble

• Title : Similarity Learning and Generalized Cosine
• Summary : Having a measure of the similarity between each pair of objects in our datasets is necessary in numerous machine learning tasks. It is the first step when using the classical supervised algorithm k-NN, or the unsupervised algorithm k-means. We will finally have a look at similarity learning in high dimensional spaces, where the classical similarity (or distance) measure is not always relevant.

— 3:30 PM Mathieu Guillame-bert, Ph.d Student at LIG Laboratory in PRIMA Team INRIA

• Title : Symbolic Temporal Data Mining
• Summary : Temporal data mining is concerned with the extraction of temporal patterns from large data sets. The usefulness of temporal data mining continues to grow as increasing amounts of temporal data about every day activities become available (recognition and prediction of person activities, fails prevision and detection, automatic Forex trading, etc.). We introduce a new temporal rule model. This model can expresses both uncertainty and temporal inaccuracy. We also present an algorithm able to extract such patterns from temporal databases. By opposition to usual approaches, this algorithm does not rely on the "windows discretization" preprocessing technique, and therefore produces patterns with high temporal accuracy.

— 4:00 PM Dimitri Masson, Ph.d Student at LIG Laboratory in PRIMA & IIHM Teams

• Title : Magellan, Using Genetic Algorithm For the Generation of User Interface
• Summary : Creativity is a major aspect in user interfaces design. While many methods and tools were devised to support creativity, examples galleries remains to be explored.This work presents Magellan, a tool that relies on Model-Driven Engineering and interactive Genetic Algorithm to automatically generate and browse galleries of examples for a particular design problem.

Opening Talk

— 4:30 PM Dimitri Masson, Ph.d Student at LIG Laboratory in PRIMA & IIHM Teams

• Title : About the Research @Stanford
• Summary : The speaker will share a personal experience on a recent visit to the Stanford HCI group. The presentation will focus on his view of the research in this prestigious research group, and in particular the "think excellent" state of mind and its implication on research and publication.

Final Discussions

— 5:00 PM Cloture & Refreshments

Objectives :
The objectives are threefold. First, the aim is to foster the interaction between PhD students of the LIG laboratory. Students will have the opportunity to exchange about their research and to meet other students and experts in the domain. Second, the objective is to create a synergy between the members of the LIG laboratory around scientific topics that are addressed by many teams. Lastly, we seek to provide LIG members with opening discussions about various topics such as education, research, industry etc. For this edition, Eustache Diemert will present how Machine Learning is used in the industry and Dimitri Masson will share a personal experience on his recent visit to the Stanford HCI group and talk in particular about the "think excellent" state of mind and its implication on research and publications.