References of "Wehenkel, Louis"
     in
Bookmark and Share    
Full Text
Peer Reviewed
See detailDecoding spontaneous brain activity from fMRI using Gaussian Processes: tracking brain reactivation
Schrouff, Jessica ULg; Kussé, Caroline ULg; Wehenkel, Louis ULg et al

in 2012 Second International Workshop on Pattern Recognition in NeuroImaging (PRNI 2012): proceedings (2012, July 03)

While Multi-Variate Pattern Analysis techniques based on machine learning have now been regularly applied to neuroimaging data, decoding brain activity is usually performed in highly controlled ... [more ▼]

While Multi-Variate Pattern Analysis techniques based on machine learning have now been regularly applied to neuroimaging data, decoding brain activity is usually performed in highly controlled experimental paradigms. In more realistic conditions, the number, sequence and duration of mental states are unpredictably generated by the individual, resulting in complex and imbalanced fMRI data sets. Moreover, in the case of spontaneous brain activity, the mental states can not be linked to any external or internal stimulation, which makes it a highly difficult condition to decode. This study tests the classification of brain activity, acquired on 14 volunteers using fMRI, during mental imagery, a condition in which the number and duration of mental events were not externally imposed but self-generated. Application of the obtained model on rest sessions allowed classifying spontaneous brain activity linked to the task which, overall, correlated with their behavioural performance to the task. [less ▲]

Detailed reference viewed: 31 (13 ULg)
Full Text
See detailL1-based compression of random forest models
Joly, Arnaud ULg; Schnitzler, François ULg; Geurts, Pierre ULg et al

in Proceeding of the 21st Belgian-Dutch Conference on Machine Learning (2012, May 24)

Random forests are effective supervised learning methods applicable to large-scale datasets. However, the space complexity of tree ensembles, in terms of their total number of nodes, is often prohibitive ... [more ▼]

Random forests are effective supervised learning methods applicable to large-scale datasets. However, the space complexity of tree ensembles, in terms of their total number of nodes, is often prohibitive, specially in the context of problems with very high-dimensional input spaces. We propose to study their compressibility by applying a L1-based regularization to the set of indicator functions defined by all their nodes. We show experimentally that preserving or even improving the model accuracy while significantly reducing its space complexity is indeed possible. [less ▲]

Detailed reference viewed: 179 (50 ULg)
Full Text
Peer Reviewed
See detailApproximation efficace de mélanges bootstrap d’arbres de Markov pour l’estimation de densité
Schnitzler, François ULg; Ammar, Sourour; Leray, Philippe et al

in Bougrain, Laurent (Ed.) Actes de la 14e Conférence Francophone sur l'Apprentissage Automatique (CAp 2012) (2012, May 23)

Nous considérons des algorithmes pour apprendre des Mélanges bootstrap d'Arbres de Markov pour l'estimation de densité. Pour les problèmes comportant un grand nombre de variables et peu d'observations ... [more ▼]

Nous considérons des algorithmes pour apprendre des Mélanges bootstrap d'Arbres de Markov pour l'estimation de densité. Pour les problèmes comportant un grand nombre de variables et peu d'observations, ces mélanges estiment généralement mieux la densité qu'un seul arbre appris au maximum de vraisemblance, mais sont plus coûteux à apprendre. C'est pourquoi nous étudions ici un algorithme pour apprendre ces modèles de manière approchée, afin d'accélérer l'apprentissage sans sacrifier la précision. Plus spécifiquement, nous récupérons lors du calcul d'un premier arbre de Markov les arcs qui constituent de bons candidats pour la structure, et ne considérons que ceux-ci lors de l'apprentissage des arbres suivants. Nous comparons cet algorithme à l'algorithme original de mélange, à un arbre appris au maximum de vraisemblance, à un arbre régularisé et à une autre méthode approchée. [less ▲]

Detailed reference viewed: 35 (4 ULg)
Full Text
Peer Reviewed
See detailSupervised learning to tune simulated annealing for in silico protein structure prediction
Marcos Alvarez, Alejandro ULg; Maes, Francis ULg; Wehenkel, Louis ULg

in Verleysen, Michel (Ed.) ESANN 2012 proceedings, 20th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (2012, April 25)

Simulated annealing is a widely used stochastic optimization algorithm whose efficiency essentially depends on the proposal distribu- tion used to generate the next search state at each step. We propose ... [more ▼]

Simulated annealing is a widely used stochastic optimization algorithm whose efficiency essentially depends on the proposal distribu- tion used to generate the next search state at each step. We propose to adapt this distribution to a family of parametric optimization problems by using supervised machine learning on a sample of search states derived from a set of typical runs of the algorithm over this family. We apply this idea in the context of in silico protein structure prediction. [less ▲]

Detailed reference viewed: 88 (32 ULg)
Full Text
Peer Reviewed
See detailStatistical interpretation of machine learning-based feature importance scores for biomarker discovery
Huynh-Thu, Vân Anh ULg; Saeys, Yvan; Wehenkel, Louis ULg et al

in Bioinformatics (2012), 28(13), 1766-1774

Motivation: Univariate statistical tests are widely used for biomarker discovery in bioinformatics. These procedures are simple, fast and their output is easily interpretable by biologists but they can ... [more ▼]

Motivation: Univariate statistical tests are widely used for biomarker discovery in bioinformatics. These procedures are simple, fast and their output is easily interpretable by biologists but they can only identify variables that provide a significant amount of information in isolation from the other variables. As biological processes are expected to involve complex interactions between variables, univariate methods thus potentially miss some informative biomarkers. Variable relevance scores provided by machine learning techniques, however, are potentially able to highlight multivariate interacting effects, but unlike the p-values returned by univariate tests, these relevance scores are usually not statistically interpretable. This lack of interpretability hampers the determination of a relevance threshold for extracting a feature subset from the rankings and also prevents the wide adoption of these methods by practicians. Results: We evaluated several, existing and novel, procedures that extract relevant features from rankings derived from machine learning approaches. These procedures replace the relevance scores with measures that can be interpreted in a statistical way, such as p-values, false discovery rates, or family wise error rates, for which it is easier to determine a significance level. Experiments were performed on several artificial problems as well as on real microarray datasets. Although the methods differ in terms of computing times and the tradeoff, they achieve in terms of false positives and false negatives, some of them greatly help in the extraction of truly relevant biomarkers and should thus be of great practical interest for biologists and physicians. As a side conclusion, our experiments also clearly highlight that using model performance as a criterion for feature selection is often counter-productive. [less ▲]

Detailed reference viewed: 159 (35 ULg)
Full Text
Peer Reviewed
See detailL1-based compression of random forest models
Joly, Arnaud ULg; Schnitzler, François ULg; Geurts, Pierre ULg et al

in 20th European Symposium on Artificial Neural Networks (2012, April)

Random forests are effective supervised learning methods applicable to large-scale datasets. However, the space complexity of tree ensembles, in terms of their total number of nodes, is often prohibitive ... [more ▼]

Random forests are effective supervised learning methods applicable to large-scale datasets. However, the space complexity of tree ensembles, in terms of their total number of nodes, is often prohibitive, specially in the context of problems with very high-dimensional input spaces. We propose to study their compressibility by applying a L1-based regularization to the set of indicator functions defined by all their nodes. We show experimentally that preserving or even improving the model accuracy while significantly reducing its space complexity is indeed possible. [less ▲]

Detailed reference viewed: 270 (65 ULg)
Full Text
Peer Reviewed
See detailLearning to play K-armed bandit problems
Maes, Francis ULg; Wehenkel, Louis ULg; Ernst, Damien ULg

in Proceedings of the 4th International Conference on Agents and Artificial Intelligence (ICAART 2012) (2012, February)

We propose a learning approach to pre-compute K-armed bandit playing policies by exploiting prior information describing the class of problems targeted by the player. Our algorithm first samples a set of K ... [more ▼]

We propose a learning approach to pre-compute K-armed bandit playing policies by exploiting prior information describing the class of problems targeted by the player. Our algorithm first samples a set of K-armed bandit problems from the given prior, and then chooses in a space of candidate policies one that gives the best average performances over these problems. The candidate policies use an index for ranking the arms and pick at each play the arm with the highest index; the index for each arm is computed in the form of a linear combination of features describing the history of plays (e.g., number of draws, average reward, variance of rewards and higher order moments), and an estimation of distribution algorithm is used to determine its optimal parameters in the form of feature weights. We carry out simulations in the case where the prior assumes a fixed number of Bernoulli arms, a fixed horizon, and uniformly distributed parameters of the Bernoulli arms. These simulations show that learned strategies perform very well with respect to several other strategies previously proposed in the literature (UCB1, UCB2, UCB-V, KL-UCB and $\epsilon_n$-GREEDY); they also highlight the robustness of these strategies with respect to wrong prior information. [less ▲]

Detailed reference viewed: 110 (19 ULg)
Full Text
Peer Reviewed
See detailDecoding Semi-Constrained Brain Activity from fMRI Using Support Vector Machines and Gaussian Processes
Schrouff, Jessica ULg; Kussé, Caroline ULg; Wehenkel, Louis ULg et al

in PLoS ONE (2012), 7(4),

Predicting a particular cognitive state from a specific pattern of fMRI voxel values is still a methodological challenge. Decoding brain activity is usually performed in highly controlled experimental ... [more ▼]

Predicting a particular cognitive state from a specific pattern of fMRI voxel values is still a methodological challenge. Decoding brain activity is usually performed in highly controlled experimental paradigms characterized by a series of distinct states induced by a temporally constrained experimental design. In more realistic conditions, the number, sequence and duration of mental states are unpredictably generated by the individual, resulting in complex and imbalanced fMRI data sets. This study tests the classification of brain activity, acquired on 16 volunteers using fMRI, during mental imagery, a condition in which the number and duration of mental events were not externally imposed but self-generated. To deal with these issues, two classification techniques were considered (Support Vector Machines, SVM, and Gaussian Processes, GP), as well as different feature extraction methods (General Linear Model, GLM and SVM). These techniques were combined in order to identify the procedures leading to the highest accuracy measures. Our results showed that 12 data sets out of 16 could be significantly modeled by either SVM or GP. Model accuracies tended to be related to the degree of imbalance between classes and to task performance of the volunteers. We also conclude that the GP technique tends to be more robust than SVM to model unbalanced data sets. [less ▲]

Detailed reference viewed: 48 (16 ULg)
Full Text
See detailAn Efficient Algorithm to Perform Multiple Testing in Epistasis Screening
Van Lishout, François ULg; Cattaert, Tom ULg; Mahachie John, Jestinah ULg et al

Conference (2011, December 13)

Background: Research in epistasis or gene-gene interaction detection for human complex traits has grown exponentially over the last few years. It has been marked by promising methodological developments ... [more ▼]

Background: Research in epistasis or gene-gene interaction detection for human complex traits has grown exponentially over the last few years. It has been marked by promising methodological developments, improved translation efforts of statistical epistasis to biological epistasis and attempts to integrate different omics information sources into the epistasis screening to enhance power. The quest for gene-gene interactions poses severe multiple-testing problems. In this context, the maxT algorithm is one technique to control the false-positive rate. However, the memory needed by this algorithm rises linearly with the amount of hypothesis tests. In main-effects detection, this is not a problem since the memory required is thus proportional to the number of SNPs. In contrast, gene-gene interaction studies will require a memory proportional to the squared amount of SNPs. A genome wide epistasis would therefore require terabytes of memory. Hence, cache problems are likely to occur, increasing the computation time. Methods: In this work we present a new version of maxT, requiring an amount of memory independent from the number of genetic effects to be investigated. This algorithm was implemented in C++ in our epistasis screening software MB-MDR-2.6.2 and compared to MB-MDR's first implementation as an R-package (Calle et al., Bioinformatics 2010). We evaluate the new implementation in terms of memory efficiency and speed using simulated data. The software is illustrated on real-life data for Crohn's disease. Results: The sequential version of MBMDR-2.6.2 is approximately 5,500 times faster than its R counterparts. The parallel version (tested on a cluster composed of 14 blades, containing each 4 quad-cores Intel Xeon CPU E5520@2.27 GHz) is approximately 900,000 times faster than the latter, for results of the same quality on the simulated data. It analyses all gene-gene interactions of a dataset of 100,000 SNPs typed on 1000 individuals within 4 days. Our program found 14 SNP-SNP interactions with a p-value less than 0.05 on the real-life Crohn’s disease data. Conclusions: Our software is able to solve large-scale SNP-SNP interactions problems within a few days, without using much memory. A new implementation to reach genome wide epistasis screening is under construction. In the context of Crohn's disease, MBMDR-2.6.2 found signal in regions well known in the field and our results could be explained from a biological point of view. This demonstrates the power of our software to find relevant phenotype-genotype associations. [less ▲]

Detailed reference viewed: 58 (25 ULg)
Full Text
See detailPruning randomized trees with L1-norm regularization
Joly, Arnaud ULg; Schnitzler, François ULg; Geurts, Pierre ULg et al

Poster (2011, November 29)

Growing amount of high dimensional data requires robust analysis techniques. Tree-based ensemble methods provide such accurate supervised learning models. However, the model complexity can become utterly ... [more ▼]

Growing amount of high dimensional data requires robust analysis techniques. Tree-based ensemble methods provide such accurate supervised learning models. However, the model complexity can become utterly huge depending on the dimension of the dataset. Here we propose a method to compress such ensemble using random tree induced space and L1-norm regularisation. This leads to a drastic pruning, preserving or improving the model accuracy. Moreover, our approach increases robustness with respect to the selection of complexity parameters. [less ▲]

Detailed reference viewed: 61 (19 ULg)
See detailDecoding semi-constrained brain activity from fMRI using SVM and GP
Schrouff, Jessica ULg; Kussé, Caroline ULg; Wehenkel, Louis ULg et al

Scientific conference (2011, November 22)

Predicting a particular cognitive state from a specific pattern of fMRI voxel values is still a methodological challenge. Decoding brain activity is usually performed in highly controlled experimental ... [more ▼]

Predicting a particular cognitive state from a specific pattern of fMRI voxel values is still a methodological challenge. Decoding brain activity is usually performed in highly controlled experimental paradigms characterized by a series of distinct states induced by a temporally constrained experimental design. In more realistic conditions, the number, sequence and duration of mental states are unpredictably generated by the individual, resulting in complex and imbalanced fMRI data sets. This study tests the classification of brain activity, acquired on 16 volunteers using fMRI, during mental imagery, a condition in which the number and duration of mental events were not externally imposed but self-generated. To deal with these issues, two classification techniques were considered (Support Vector Machines, SVM, and Gaussian Processes, GP), as well as different feature extraction methods (General Linear Model, GLM and SVM). These techniques were combined in order to identify the procedures leading to the highest accuracy measures. Our results showed that 12 data sets out of 16 could be significantly modeled by either SVM or GP. Model accuracies tended to be related to the degree of imbalance between classes and to task performance of the volunteers. We also conclude that the GP technique tends to be more robust than SVM to model unbalanced data sets. [less ▲]

Detailed reference viewed: 51 (5 ULg)
Full Text
Peer Reviewed
See detailPhenotype Classification of Zebrafish Embryos by Supervised Learning
Jeanray, Nathalie ULg; Marée, Raphaël ULg; Pruvot, Benoist ULg et al

Conference (2011, September 02)

Detailed reference viewed: 34 (13 ULg)
Full Text
Peer Reviewed
See detailDistributed MPC of wide-area electromechanical oscillations of large-scale power systems
Wang, Da ULg; Glavic, Mevludin; Wehenkel, Louis ULg

in Proceedings of ISAP 2011 (2011, September)

We investigate distributed Model Predictive Control (MPC) to damp wide-area electromechanical oscillations. Our distributed MPC schemes are derived from and compared with a fully centralized MPC scheme ... [more ▼]

We investigate distributed Model Predictive Control (MPC) to damp wide-area electromechanical oscillations. Our distributed MPC schemes are derived from and compared with a fully centralized MPC scheme proposed in a previous publication. Based on simulations carried out using a 16-generator, 70-bus, two-area test power system, we show that simple coordination schemes based on additional local measurements’ feedback yield already a significant improvement with respect to a scheme with only implicit coordination, improve significantly with respect to purely local controls, and in this respect reach about 75% of the improvements obtained by an ideal centralized MPC scheme. [less ▲]

Detailed reference viewed: 32 (7 ULg)
Full Text
Peer Reviewed
See detailEfficiently approximating Markov tree bagging for high-dimensional density estimation
Schnitzler, François ULg; ammar, sourour; leray, philippe et al

in Gunopulos, Dimitrios; Hofmann, Thomas; Malerba, Donato (Eds.) et al Machine Learning and Knowledge Discovery in Databases, Part III (2011, September)

We consider algorithms for generating Mixtures of Bagged Markov Trees, for density estimation. In problems defined over many variables and when few observations are available, those mixtures generally ... [more ▼]

We consider algorithms for generating Mixtures of Bagged Markov Trees, for density estimation. In problems defined over many variables and when few observations are available, those mixtures generally outperform a single Markov tree maximizing the data likelihood, but are far more expensive to compute. In this paper, we describe new algorithms for approximating such models, with the aim of speeding up learning without sacrificing accuracy. More specifically, we propose to use a filtering step obtained as a by-product from computing a first Markov tree, so as to avoid considering poor candidate edges in the subsequently generated trees. We compare these algorithms (on synthetic data sets) to Mixtures of Bagged Markov Trees, as well as to a single Markov tree derived by the classical Chow-Liu algorithm and to a recently proposed randomized scheme used for building tree mixtures. [less ▲]

Detailed reference viewed: 73 (23 ULg)
Full Text
Peer Reviewed
See detailDay-ahead Security Assessment under Uncertainty Relying on the Combination of Preventive and Corrective Controls to Face Worst-Case Scenarios
Capitanescu, Florin ULg; Fliscounakis, Stéphane; Panciatici, Patrick et al

in PSCC proceedings Stockholm (Sweden) 2011 (2011, August 22)

This paper deals with day-ahead static security assessment with respect to a postulated set of contingencies while taking into account uncertainties about the next day system conditions. We propose a ... [more ▼]

This paper deals with day-ahead static security assessment with respect to a postulated set of contingencies while taking into account uncertainties about the next day system conditions. We propose a heuristic approach to check whether, given some assumptions regarding these uncertainties, the worst case with respect to each contingency is still controllable by appropriate combinations of preventive and corrective actions. This approach relies on the solution of successive optimal power flow (OPF) and security-constrained optimal power flow (SCOPF) problems of a special type. The interest of the approach is shown by illustrative examples on the Nordic32 system. [less ▲]

Detailed reference viewed: 145 (6 ULg)
Full Text
Peer Reviewed
See detailState-of-the-art, challenges, and future trends in security constrained optimal power flow
Capitanescu, Florin ULg; Martinez Ramos, Jose Luis; Panciatici, Patrick et al

in Electric Power Systems Research (2011), 81(8), 1731-1741

This paper addresses the main challenges to the security constrained optimal power flow (SCOPF) computations. We first discuss the issues related to the SCOPF problem formulation such as the use of a ... [more ▼]

This paper addresses the main challenges to the security constrained optimal power flow (SCOPF) computations. We first discuss the issues related to the SCOPF problem formulation such as the use of a limited number of corrective actions in the post-contingency states and the modeling of voltage and transient stability constraints. Then we deal with the challenges to the techniques for solving the SCOPF, focusing mainly on: approaches to reduce the size of the problem by either efficiently identifying the binding contingencies and including only these contingencies in the SCOPF or by using approximate models for the post-contingency states, and the handling of discrete variables. We finally address the current trend of extending the SCOPF formulation to take into account the increasing levels of uncertainty in the operation planning. For each such topic we provide a review of the state of the art, we identify the advances that are needed, and we indicate ways to bridge the gap between the current state of the art and these needs. [less ▲]

Detailed reference viewed: 259 (15 ULg)
Full Text
Peer Reviewed
See detailSOLVING VERY LARGE-SCALE SECURITY-CONSTRAINED OPTIMAL POWER FLOW PROBLEMS BY COMBINING ITERATIVE CONTINGENCY SELECTION AND NETWORK COMPRESSION
Platbrood, Ludovic; Crisciu, Horia; Capitanescu, Florin ULg et al

in PSCC conference (2011, August)

This paper proposes a practical algorithm for solving very large-scale SCOPF problems, based on the combination of a contingency filtering scheme, used to identify the binding contingencies at the optimum ... [more ▼]

This paper proposes a practical algorithm for solving very large-scale SCOPF problems, based on the combination of a contingency filtering scheme, used to identify the binding contingencies at the optimum, and a network compression method, used to reduce the complexity of the post-contingency models included in the SCOPF formulation. By combining these two complementary simplifications, it is possible to solve SCOPF problems addressing both preventive and corrective controls on continental sized power system models and with a very large number of contingencies. The proposed algorithms are implemented with state-of-the-art solvers and applied on a model of the European transmission system, of about 15000 buses, and with about 11000 contingencies. [less ▲]

Detailed reference viewed: 158 (1 ULg)
Full Text
Peer Reviewed
See detailRedispatching active and reactive powers using a limited number of control actions
Capitanescu, Florin ULg; Wehenkel, Louis ULg

in IEEE Transactions on Power Systems (2011), 26(3), 1221-1230

This paper deals with some essential open questions in the field of optimal power flow (OPF) computations, namely: the limitation of the number of controls allowed to move, the trade-off between the ... [more ▼]

This paper deals with some essential open questions in the field of optimal power flow (OPF) computations, namely: the limitation of the number of controls allowed to move, the trade-off between the objective function and the number of controls allowed to move, the computation of the minimum number of control actions needed to satisfy constraints, and the determination of the sequence of control actions to be taken by the system operator in order to achieve its operation goal. To address these questions, we propose approaches which rely on the computation of sensitivities of the objective function and inequality constraints with respect to control actions. We thus determine a subset of controls allowed to move in the OPF, by solving a sensitivity-based mixed integer linear programming (MILP) problem. We study the performances of these approaches on three test systems (of 60, 118, and 618 buses) and by considering three different OPF problems important for a system operator in emergency and/or in normal states, namely the removal of thermal congestions, the removal of bus voltage limits violation, and the reduction of the active power losses. [less ▲]

Detailed reference viewed: 123 (12 ULg)