You are here

News

  • MLG seminar - Introduction to convolutional neural networks

    On September 16, 2015 a seminar will be presented by Laura Mannie-Corbisier with the title "Introduction to convolutional neural networks". This seminar will take place on the ULB campus, building NO, 9th floor, local NO9.06 at 11:00.

  • MLG seminar - Just In Time Classifiers For Recurrent Concepts

    On September 16, 2015 a seminar will be presented by Giacomo Boracchi with the title "Just In Time Classifiers For Recurrent Concepts". This seminar will take place on the ULB campus, building NO, 9th floor, local NO9.06 at 12PM.

  • [PhD public defense] Inference of gene networks from time series expression data and application to type 1 Diabetes

    Public defense of the PhD thesis of Miguel Lopes entitled "Inference of gene networks from time series expression data and application to type 1 Diabetes". The defence will take place at the Université libre de Bruxelles (in the La Plaine, Forum G) on Friday, September 4, at 4PM. Abstract: The inference of gene regulatory networks (GRN) is of great importance to medical research, as causal mechanisms responsible for phenotypes are unravelled and potential therapeutical targets identified. In type 1 diabetes, insulin producing pancreatic beta-cells are the target of an auto-immune attack leading to apoptosis (cell suicide). Although key genes and regulations have been identified, a precise characterization of the process leading to beta-cell apoptosis has not been achieved yet. The inference of relevant molecular pathways in type 1 diabetes is then a crucial research topic. GRN inference from gene expression data may be tackled with well-established statistical and machine learning tools. In particular, the use of time series facilitates the identification of the causal direction in cause-effect gene pairs. GRN inference is a very challenging task due to the very high number of genes and typical low number of available samples. The first part of this presentation will present novel heuristics to GRN inference from time series, designed to deal with the high variable to sample ratio. State of the art approaches are described and assessed in real and simulated data. The second part of the presentation is on the context of type 1 diabetes, and consists of a study on beta cell gene expression after exposure to cytokines, emulating the mechanisms leading to apoptosis. Multiple datasets of beta cell gene expression were used to identify differentially expressed genes, and a regulatory network involving them was inferred. Top differentially expressed genes were found to modulate cytokine induced apoptosis and predicted regulations were experimentally confirmed.

  • MLG seminar - When is undersampling effective in unbalanced classification tasks?

    On September 2, 2015 a seminar will be presented by Andrea Dal Pozzolo with the title "When is undersampling effective in unbalanced classification tasks?". This seminar will take place on the ULB campus, building NO, 8th floor, P.2NO8.08 at 4PM.

  • MLG co-organizes AAAI Spring Symposium at Stanford University

    Tom Lenaerts co-organizes with Bipin Indurkhya, Georgi Stojanov, Joanna Bryson and Tony Vaele an AAAI Spring Symposium on ethical and moral considerations in non-human agents  at Stanford University from March 21-23, 2016.The deadline for submitting a 2 page abstract is October 5th 2015.More details concerning this AAAI Sping Symposium can be found on https://sites.google.com/site/ethicalnonhumanagents/ 

  • MLG Fulbright student wins RCSB PDB Poster prize at ISMB/ECCB 2015

    Ashley Conard, Fulbright and BAEF student working with Elisa Cilia and Tom Lenaerts, won the RCSB PDB poster prize at ISMB/ECCB 2015.  More information can be found on : https://www.iscb.org/ismbeccb2015-program/award-winners   

  • MLG seminar - Context-sensitive Ordinal Regression Models for Human Facial Behaviour Analysis

    On July 8, 2015 a seminar will be presented by Ognjen Rudovic with the title "Context-sensitive Ordinal Regression Models for Human Facial Behaviour Analysis". This seminar will take place on the ULB campus, building NO, 8th floor, P.2NO8.08 at 11:30.

  • New MLG publication in Nature's Scientific Reports

    Why do we apologize? Why do we forgive? Would it not simply be better to end a relationship that is failing, even when it has been and still can be beneficial?  In a mathematical investigation, Luis Martinez-Vaquero and Tom Lenaerts of MLG show, in collaboration with international colleagues The Anh Han (U. Teesside, UK) and Luís Moniz Pereira (UNL, Portugal), that revenge, apology and forgiveness may actually have been shaped by evolution to deal with just such situations. Moreover, analysis shows that an apology actually needs to be sincere as otherwise the mechanism for sustained cooperation fails, leading either to revenge or to non-cooperative strategies. This work has been now reported in Nature’s multi-disciplinary open access journal Scientific Reports.  Article available via website Scientific Reports

  • [PhD Public defense] A Machine Learning Approach for Automatic and Generic Side-Channel Attacks

    Public defense of the PhD thesis of Liran Lerman entitled "A Machine Learning Approach for Automatic and Generic Side-Channel Attacks" and supervised by Prof. Gianluca Bontempi and Prof. Olivier Markowitch. The defence (in french) will take place at the Université libre de Bruxelles (in the La Plaine, N-O building, 5th floor, Salle Solvay) on Wednesday, June 10, at 16:30. Abstract: The pervasive presence of interconnected devices has lead to a massive interest in security features provided among others by cryptography. For decades, designers estimated the security level of a cryptographic algorithm independently of its implementation in a cryptographic device viagra for sale uk. However, since the publication on implementation attacks in 1996, the physical attacks have become an active research area by analysing physical leakages measured on the target cryptographic device. In this work, we focus on profiled attacks that were introduced as the most efficient class of strategies in order to evaluate the physical leakages of cryptographic devices. The traditional profiled attacks apply parametric methods in which a priori information on the physical properties is assumed. Machine learning represents an active research area that allows producing automatic and generic models requiring no a priori information on the underlying phenomenon. The purpose of this presentation is to shed new light on the capabilities of learning models for physical analysis.

Pages

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer