Stéphane Canu | INSA ROUEN (original) (raw)
Papers by Stéphane Canu
The Journal of Machine Learning Research, Dec 1, 2005
This work deals with a method for building a reproducing kernel Hilbert space (RKHS) from a Hilbe... more This work deals with a method for building a reproducing kernel Hilbert space (RKHS) from a Hilbert space with frame elements having special properties. Conditions on existence and a method of construction are given. Then, these RKHS are used within the framework of regularization theory for function approximation. Implications on semiparametric estimation are discussed and a multiscale scheme of regularization is also proposed. Results on toy and real-world approximation problems illustrate the ...
IEEE Transactions on Neural Networks and Learning Systems, 2014
Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several pr... more Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several preprocessing approaches have been proposed, only a few works have been focused on integrating the feature selection into the learning process. In this work, we propose a general framework for feature selection in learning to rank using SVM with a sparse regularization term. We investigate both classical convex regularizations such as ℓ1 or weighted ℓ1 and non-convex regularization terms such as log penalty, Minimax Concave Penalty (MCP) or ℓp pseudo norm with p < 1. Two algorithms are proposed, first an accelerated proximal approach for solving the convex problems, second a reweighted ℓ1 scheme to address the non-convex regularizations. We conduct intensive experiments on nine datasets from Letor 3.0 and Letor 4.0 corpora. Numerical results show that the use of non-convex regularizations we propose leads to more sparsity in the resulting models while prediction performance is preserved. The number of features is decreased by up to a factor of six compared to the ℓ1 regularization. In addition, the software is publicly available on the web. 1 2 #{(ds,dt)∈D|dt≻xds and dt≻yds} #{(ds,dt)∈D}
Proceedings Second International Conference on Cognitive Technology Humanizing the Information Age, 1997
IEEE Transactions on Image Processing, 2000
Building an accurate training database is challenging in supervised classification. For instance,... more Building an accurate training database is challenging in supervised classification. For instance, in medical imaging, radiologists often delineate malignant and benign tissues without access to the histological ground truth, leading to uncertain datasets. This paper addresses the pattern classification problem arising when available target data include some uncertainty information. Target data considered here are both qualitative (a class label) or quantitative (an estimation of the posterior probability). In this context, usual discriminative methods such as support vector machine (SVM) fail either to learn a robust classifier or to predict accurate probability estimates. We generalize the regular SVM by introducing a new formulation of the learning problem to take into account class labels as well as class probability estimates. This original reformulation into a probabilistic SVM (P-SVM) can be efficiently solved by adapting existing flexible SVM solvers. Furthermore, this framework allows deriving a unique learned prediction function for both decision and posterior probability estimation providing qualitative and quantitative predictions. The method is first tested on synthetic datasets to evaluate its properties as compared with the classical SVM and Fuzzy-SVM. It is then evaluated on a clinical dataset of multiparametric prostate magnetic resonance (mpMR) images to assess its performances in discriminating benign from malignant tissues. It is shown to outperform classical SVM in terms of probability predictions and classification performances, and demonstrates its potential for the design of an efficient computer-aided decision (CAD) systems for prostate cancer diagnosis based on multiparametric MR imaging.
IEEE Transactions on Neural Networks, 2000
Recently, there has been a lot of interest around multi-task learning (MTL) problem with the cons... more Recently, there has been a lot of interest around multi-task learning (MTL) problem with the constraints that tasks should share a common sparsity profile. Such a problem can be addressed through a regularization framework where the regularizer induces a joint-sparsity pattern between task decision functions. We follow this principled framework and focus on ℓp − ℓq (with 0 ≤ p ≤ 1 and 1 ≤ q ≤ 2) mixed-norms as sparsityinducing penalties. Our motivation for addressing such a larger class of penalty is to adapt the penalty to a problem at hand leading thus to better performances and better sparsity pattern. For solving the problem in the general multiple kernel case, we first derive a variational formulation of the ℓ1 − ℓq penalty which helps us in proposing an alternate optimization algorithm. Although very simple, the latter algorithm provably converges to the global minimum of the ℓ1 − ℓq penalized problem. For the linear case, we extend existing works considering accelerated proximal gradient to this penalty. Our contribution in this context is to provide an efficient scheme for computing the ℓ1 − ℓq proximal operator. Then, for the more general case when 0 < p < 1, we solve the resulting non-convex problem through a majorization-minimization approach. The resulting algorithm is an iterative scheme which, at each iteration, solves a weighted ℓ1 − ℓq sparse MTL problem. Empirical evidences from toy dataset and real-word datasets dealing with BCI single trial EEG classification and protein subcellular localization show the benefit of the proposed approaches and algorithms.
The Learning Workshop, Feb 19, 2009
Recently, there has been a lot of interest around multi-task learning (MTL) problem with the cons... more Recently, there has been a lot of interest around multi-task learning (MTL) problem with the constraints that tasks should share common features. Such a problem can be addressed through a regularization framework where the regularizer induces a joint-sparsity pattern between task decision functions. We follow this principled framework but instead we focus on ip− i2 (with p≤ 1) mixed-norms as sparsity-inducing penalties. After having shown that the i1− i2 MTL problem is a general case of Multiple Kernel Learning (MKL), we ...
2009 IEEE International Workshop on Machine Learning for Signal Processing, 2009
Sequence labeling is concerned with processing an input data sequence and producing an output seq... more Sequence labeling is concerned with processing an input data sequence and producing an output sequence of discrete labels which characterize it. Common applications includes speech recognition, language processing (tagging, chunking) and bioinformatics. Many solutions have been proposed to partially cope with this problem. These include probabilistic models (HMMs, CRFs) and machine learning algorithm (SVM, Neural nets). In practice, the best results have been obtained by combining several of these methods. However, fusing different signal segmentation methods is not straightforward, particularly when integrating prior information. In this paper the sequence labeling problem is viewed as a multi objective optimization task. Each objective targets a different aspect of sequence labelling such as good classification, temporal stability and change detection. The resulting optimization problem turns out to be non convex and plagued with numerous local minima. A region growing algorithm is proposed as a method for finding a solution to this multi functional optimization task. The proposed algorithm is evaluated on both synthetic and real data (BCI dataset). Results are encouraging and better than those previously reported on these datasets.
2011 IEEE Statistical Signal Processing Workshop (SSP), 2011
This paper addresses the pattern classification problem arising when available target data includ... more This paper addresses the pattern classification problem arising when available target data include some uncertainty information. Target data considered here is either qualitative (a class label) or quantitative (an estimation of the posterior probability). Our main contribution is a SVM inspired formulation of this problem allowing to take into account class label through a hinge loss as well as probability estimates using ε-insensitive cost function together with a minimum norm (maximum margin) objective. This formulation shows a dual form leading to a quadratic problem and allows the use of a representer theorem and associated kernel. The solution provided can be used for both decision and posterior probability estimation. Based on empirical evidence our method outperforms regular SVM in terms of probability predictions and classification performances.
Ecological Modelling, 1999
Software sensor design consists of building an estimate of some quantity of interest. This estima... more Software sensor design consists of building an estimate of some quantity of interest. This estimate can be used either to replace a physical measurement, or to validate an existing one. This paper provides some general guidelines for the design of software sensors based on empirical data. When the model is a priori unknown, the problem can be stated in terms of non-parametric regression or black-box modelling. Complexity control is the main difficulty in this setting. A trade-off must be achieved between two antagonist goals: the model ...
Parasitology Research, Mar 1, 2006
9th International Conference on Artificial Neural Networks: ICANN '99, 1999
Abstract The use of radial basis functions in supervised learning is well motivated by approximat... more Abstract The use of radial basis functions in supervised learning is well motivated by approximation theory. Computation issues have lead us to consider some approximations of this scheme, losing much of the mathematical foundation in the process. We show that basis pursuit denoising is a principled alternative to classical RBF, which leads to sparse expansions. This alternative is local in the sense that complexity is tuned locally. A further step in this direction is made by adapting the locality parameter of each basis function. ...
Support Vector Machines (SVM) is a new machine learning approach based on Statistical Learning Th... more Support Vector Machines (SVM) is a new machine learning approach based on Statistical Learning Theory (Vapnik-Chervonenkis or VC-theory). VC- theory has a solid mathematical background for the dependencies estimation and predictive learning from finite data sets. SVM is based on the Structural Risk Minimisation principle, aiming to minimise both the empirical risk and the complexity of the model, providing high
Revue de l'Electricité et de l'Electronique, 2006
Résumé/Abstract Cet article présente une introduction illustrée des idées qui ont permis l&am... more Résumé/Abstract Cet article présente une introduction illustrée des idées qui ont permis l&#x27;émergence d&#x27;une nouvelle classe de discriminateur universel très efficaces: les séparateurs à vaste marge (SVM). Parmi ces idées on retrouve la notion de critère à minimiser, l&#x27;utilisation de noyaux et la volonté de poser le problème formellement dès le début de l&#x27;analyse.
Lecture Notes in Computer Science, 2002
This paper deals with a way of constructing reproducing kernel Hilbert spaces and their associate... more This paper deals with a way of constructing reproducing kernel Hilbert spaces and their associated kernels from frame theory. After introducing briefly frame theory, we give mild conditions on frame elements for spanning a RKHS. Examples of different kernels are then given based on wavelet frame. Thus, issues of this way of building kernel for semiparametric learning are discussed and an application example on a toy problem is described.
International Statistical Review, 2014
In this article, we develop a modern perspective on Akaike's Information Criterion (AIC) and Mall... more In this article, we develop a modern perspective on Akaike's Information Criterion (AIC) and Mallows' C p for model selection, and proposes generalizations to spherically and elliptically symmetric distributions. Despite the differences in their respective motivation, C p and AIC are equivalent in the special case of Gaussian linear regression. In this case they are also equivalent to a third criterion, an unbiased estimator of the quadratic prediction loss, derived from loss estimation theory. We then show that the form of the unbiased estimator of the quadratic prediction loss under a Gaussian assumption still holds under a more general distributional assumption, the family of spherically symmetric distributions. One of the features of our results is that our criterion does not rely on the specificity of the distribution, but only on its spherical symmetry. The same kind of criterion can be derived for a family of elliptically contoured distribution, which allows correlations, when considering the invariant loss. More specifically, the unbiasedness property is relative to a distribution associated to the original density.
Ingénierie des systèmes d'information, 2003
Résumé/Abstract L&#x27;asthme représente une maladie chronique fréquente qui, malgré des ... more Résumé/Abstract L&#x27;asthme représente une maladie chronique fréquente qui, malgré des traitements efficaces disponibles, reste insuffisamment contrôlée. Dans ce cadre, nous proposons la mise au point d&#x27;un système d&#x27;aide à la décision pour la prise en charge de l&#x27;asthme. Notre première contribution se situe dans le choix du raisonnement à partir de cas (RàPC) comme méthodologie de résolution de problèmes. Après une brève présentation du RàPC, nous détaillons la mise au point de notre prototype ADEMA. Dans un premier lieu, ...
IEEE Transactions on Neural Networks and Learning Systems, 2014
Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several pr... more Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several preprocessing approaches have been proposed, only a few works have been focused on integrating the feature selection into the learning process. In this work, we propose a general framework for feature selection in learning to rank using SVM with a sparse regularization term. We investigate both classical convex regularizations such as ℓ1 or weighted ℓ1 and non-convex regularization terms such as log penalty, Minimax Concave Penalty (MCP) or ℓp pseudo norm with p < 1. Two algorithms are proposed, first an accelerated proximal approach for solving the convex problems, second a reweighted ℓ1 scheme to address the non-convex regularizations. We conduct intensive experiments on nine datasets from Letor 3.0 and Letor 4.0 corpora. Numerical results show that the use of non-convex regularizations we propose leads to more sparsity in the resulting models while prediction performance is preserved. The number of features is decreased by up to a factor of six compared to the ℓ1 regularization. In addition, the software is publicly available on the web. 1 2 #{(ds,dt)∈D|dt≻xds and dt≻yds} #{(ds,dt)∈D}
Journal Européen des Systèmes Automatisés, 2005
Résumé/Abstract The paper investigates the application of black box modelling to the prediction o... more Résumé/Abstract The paper investigates the application of black box modelling to the prediction of the daily maxima of ground-ozone level. The main interest of these modelling approaches is their genericity as they are solely based on the available data provided by the Associations of air quality monitoring and they can be transposed from a geographical area to another one. The paper realises a comparative study of four statistical learning approaches, the decisions trees (, the neural networks, the least-angle regression and the ...
2013 Humaine Association Conference on Affective Computing and Intelligent Interaction, 2013
ABSTRACT The objective of the paper is to investigate the effect of subject's emotional s... more ABSTRACT The objective of the paper is to investigate the effect of subject's emotional states on Brain Computer Interface (BCI) performance. Two psycho-physiological experiments are designed and implemented. The first one induces subjects' emotion using video clips first, then involves subjects' in SSVEP task. The second one induces subjects' emotions and SSVEP simultaneously by flickering IAPS pictures in four directions. used to recognize the performed BCI tasks. Based on the performances of learned classifiers, we analyzed the influence of emotion using two statistical tests. The McNamara's test serves to assess if emotion has any influences on mental task performing while Wilcox on signed-rank test analyses if emotion has a positive or detrimental effect on ability to achieve a SSVEP task. Obtained results suggest influence of emotional states: the positive and neutral emotions influence BCI performance similarly, while the negative emotion tends to deteriorate classification accuracy.
The Journal of Machine Learning Research, Dec 1, 2005
This work deals with a method for building a reproducing kernel Hilbert space (RKHS) from a Hilbe... more This work deals with a method for building a reproducing kernel Hilbert space (RKHS) from a Hilbert space with frame elements having special properties. Conditions on existence and a method of construction are given. Then, these RKHS are used within the framework of regularization theory for function approximation. Implications on semiparametric estimation are discussed and a multiscale scheme of regularization is also proposed. Results on toy and real-world approximation problems illustrate the ...
IEEE Transactions on Neural Networks and Learning Systems, 2014
Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several pr... more Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several preprocessing approaches have been proposed, only a few works have been focused on integrating the feature selection into the learning process. In this work, we propose a general framework for feature selection in learning to rank using SVM with a sparse regularization term. We investigate both classical convex regularizations such as ℓ1 or weighted ℓ1 and non-convex regularization terms such as log penalty, Minimax Concave Penalty (MCP) or ℓp pseudo norm with p < 1. Two algorithms are proposed, first an accelerated proximal approach for solving the convex problems, second a reweighted ℓ1 scheme to address the non-convex regularizations. We conduct intensive experiments on nine datasets from Letor 3.0 and Letor 4.0 corpora. Numerical results show that the use of non-convex regularizations we propose leads to more sparsity in the resulting models while prediction performance is preserved. The number of features is decreased by up to a factor of six compared to the ℓ1 regularization. In addition, the software is publicly available on the web. 1 2 #{(ds,dt)∈D|dt≻xds and dt≻yds} #{(ds,dt)∈D}
Proceedings Second International Conference on Cognitive Technology Humanizing the Information Age, 1997
IEEE Transactions on Image Processing, 2000
Building an accurate training database is challenging in supervised classification. For instance,... more Building an accurate training database is challenging in supervised classification. For instance, in medical imaging, radiologists often delineate malignant and benign tissues without access to the histological ground truth, leading to uncertain datasets. This paper addresses the pattern classification problem arising when available target data include some uncertainty information. Target data considered here are both qualitative (a class label) or quantitative (an estimation of the posterior probability). In this context, usual discriminative methods such as support vector machine (SVM) fail either to learn a robust classifier or to predict accurate probability estimates. We generalize the regular SVM by introducing a new formulation of the learning problem to take into account class labels as well as class probability estimates. This original reformulation into a probabilistic SVM (P-SVM) can be efficiently solved by adapting existing flexible SVM solvers. Furthermore, this framework allows deriving a unique learned prediction function for both decision and posterior probability estimation providing qualitative and quantitative predictions. The method is first tested on synthetic datasets to evaluate its properties as compared with the classical SVM and Fuzzy-SVM. It is then evaluated on a clinical dataset of multiparametric prostate magnetic resonance (mpMR) images to assess its performances in discriminating benign from malignant tissues. It is shown to outperform classical SVM in terms of probability predictions and classification performances, and demonstrates its potential for the design of an efficient computer-aided decision (CAD) systems for prostate cancer diagnosis based on multiparametric MR imaging.
IEEE Transactions on Neural Networks, 2000
Recently, there has been a lot of interest around multi-task learning (MTL) problem with the cons... more Recently, there has been a lot of interest around multi-task learning (MTL) problem with the constraints that tasks should share a common sparsity profile. Such a problem can be addressed through a regularization framework where the regularizer induces a joint-sparsity pattern between task decision functions. We follow this principled framework and focus on ℓp − ℓq (with 0 ≤ p ≤ 1 and 1 ≤ q ≤ 2) mixed-norms as sparsityinducing penalties. Our motivation for addressing such a larger class of penalty is to adapt the penalty to a problem at hand leading thus to better performances and better sparsity pattern. For solving the problem in the general multiple kernel case, we first derive a variational formulation of the ℓ1 − ℓq penalty which helps us in proposing an alternate optimization algorithm. Although very simple, the latter algorithm provably converges to the global minimum of the ℓ1 − ℓq penalized problem. For the linear case, we extend existing works considering accelerated proximal gradient to this penalty. Our contribution in this context is to provide an efficient scheme for computing the ℓ1 − ℓq proximal operator. Then, for the more general case when 0 < p < 1, we solve the resulting non-convex problem through a majorization-minimization approach. The resulting algorithm is an iterative scheme which, at each iteration, solves a weighted ℓ1 − ℓq sparse MTL problem. Empirical evidences from toy dataset and real-word datasets dealing with BCI single trial EEG classification and protein subcellular localization show the benefit of the proposed approaches and algorithms.
The Learning Workshop, Feb 19, 2009
Recently, there has been a lot of interest around multi-task learning (MTL) problem with the cons... more Recently, there has been a lot of interest around multi-task learning (MTL) problem with the constraints that tasks should share common features. Such a problem can be addressed through a regularization framework where the regularizer induces a joint-sparsity pattern between task decision functions. We follow this principled framework but instead we focus on ip− i2 (with p≤ 1) mixed-norms as sparsity-inducing penalties. After having shown that the i1− i2 MTL problem is a general case of Multiple Kernel Learning (MKL), we ...
2009 IEEE International Workshop on Machine Learning for Signal Processing, 2009
Sequence labeling is concerned with processing an input data sequence and producing an output seq... more Sequence labeling is concerned with processing an input data sequence and producing an output sequence of discrete labels which characterize it. Common applications includes speech recognition, language processing (tagging, chunking) and bioinformatics. Many solutions have been proposed to partially cope with this problem. These include probabilistic models (HMMs, CRFs) and machine learning algorithm (SVM, Neural nets). In practice, the best results have been obtained by combining several of these methods. However, fusing different signal segmentation methods is not straightforward, particularly when integrating prior information. In this paper the sequence labeling problem is viewed as a multi objective optimization task. Each objective targets a different aspect of sequence labelling such as good classification, temporal stability and change detection. The resulting optimization problem turns out to be non convex and plagued with numerous local minima. A region growing algorithm is proposed as a method for finding a solution to this multi functional optimization task. The proposed algorithm is evaluated on both synthetic and real data (BCI dataset). Results are encouraging and better than those previously reported on these datasets.
2011 IEEE Statistical Signal Processing Workshop (SSP), 2011
This paper addresses the pattern classification problem arising when available target data includ... more This paper addresses the pattern classification problem arising when available target data include some uncertainty information. Target data considered here is either qualitative (a class label) or quantitative (an estimation of the posterior probability). Our main contribution is a SVM inspired formulation of this problem allowing to take into account class label through a hinge loss as well as probability estimates using ε-insensitive cost function together with a minimum norm (maximum margin) objective. This formulation shows a dual form leading to a quadratic problem and allows the use of a representer theorem and associated kernel. The solution provided can be used for both decision and posterior probability estimation. Based on empirical evidence our method outperforms regular SVM in terms of probability predictions and classification performances.
Ecological Modelling, 1999
Software sensor design consists of building an estimate of some quantity of interest. This estima... more Software sensor design consists of building an estimate of some quantity of interest. This estimate can be used either to replace a physical measurement, or to validate an existing one. This paper provides some general guidelines for the design of software sensors based on empirical data. When the model is a priori unknown, the problem can be stated in terms of non-parametric regression or black-box modelling. Complexity control is the main difficulty in this setting. A trade-off must be achieved between two antagonist goals: the model ...
Parasitology Research, Mar 1, 2006
9th International Conference on Artificial Neural Networks: ICANN '99, 1999
Abstract The use of radial basis functions in supervised learning is well motivated by approximat... more Abstract The use of radial basis functions in supervised learning is well motivated by approximation theory. Computation issues have lead us to consider some approximations of this scheme, losing much of the mathematical foundation in the process. We show that basis pursuit denoising is a principled alternative to classical RBF, which leads to sparse expansions. This alternative is local in the sense that complexity is tuned locally. A further step in this direction is made by adapting the locality parameter of each basis function. ...
Support Vector Machines (SVM) is a new machine learning approach based on Statistical Learning Th... more Support Vector Machines (SVM) is a new machine learning approach based on Statistical Learning Theory (Vapnik-Chervonenkis or VC-theory). VC- theory has a solid mathematical background for the dependencies estimation and predictive learning from finite data sets. SVM is based on the Structural Risk Minimisation principle, aiming to minimise both the empirical risk and the complexity of the model, providing high
Revue de l'Electricité et de l'Electronique, 2006
Résumé/Abstract Cet article présente une introduction illustrée des idées qui ont permis l&am... more Résumé/Abstract Cet article présente une introduction illustrée des idées qui ont permis l&#x27;émergence d&#x27;une nouvelle classe de discriminateur universel très efficaces: les séparateurs à vaste marge (SVM). Parmi ces idées on retrouve la notion de critère à minimiser, l&#x27;utilisation de noyaux et la volonté de poser le problème formellement dès le début de l&#x27;analyse.
Lecture Notes in Computer Science, 2002
This paper deals with a way of constructing reproducing kernel Hilbert spaces and their associate... more This paper deals with a way of constructing reproducing kernel Hilbert spaces and their associated kernels from frame theory. After introducing briefly frame theory, we give mild conditions on frame elements for spanning a RKHS. Examples of different kernels are then given based on wavelet frame. Thus, issues of this way of building kernel for semiparametric learning are discussed and an application example on a toy problem is described.
International Statistical Review, 2014
In this article, we develop a modern perspective on Akaike's Information Criterion (AIC) and Mall... more In this article, we develop a modern perspective on Akaike's Information Criterion (AIC) and Mallows' C p for model selection, and proposes generalizations to spherically and elliptically symmetric distributions. Despite the differences in their respective motivation, C p and AIC are equivalent in the special case of Gaussian linear regression. In this case they are also equivalent to a third criterion, an unbiased estimator of the quadratic prediction loss, derived from loss estimation theory. We then show that the form of the unbiased estimator of the quadratic prediction loss under a Gaussian assumption still holds under a more general distributional assumption, the family of spherically symmetric distributions. One of the features of our results is that our criterion does not rely on the specificity of the distribution, but only on its spherical symmetry. The same kind of criterion can be derived for a family of elliptically contoured distribution, which allows correlations, when considering the invariant loss. More specifically, the unbiasedness property is relative to a distribution associated to the original density.
Ingénierie des systèmes d'information, 2003
Résumé/Abstract L&#x27;asthme représente une maladie chronique fréquente qui, malgré des ... more Résumé/Abstract L&#x27;asthme représente une maladie chronique fréquente qui, malgré des traitements efficaces disponibles, reste insuffisamment contrôlée. Dans ce cadre, nous proposons la mise au point d&#x27;un système d&#x27;aide à la décision pour la prise en charge de l&#x27;asthme. Notre première contribution se situe dans le choix du raisonnement à partir de cas (RàPC) comme méthodologie de résolution de problèmes. Après une brève présentation du RàPC, nous détaillons la mise au point de notre prototype ADEMA. Dans un premier lieu, ...
IEEE Transactions on Neural Networks and Learning Systems, 2014
Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several pr... more Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several preprocessing approaches have been proposed, only a few works have been focused on integrating the feature selection into the learning process. In this work, we propose a general framework for feature selection in learning to rank using SVM with a sparse regularization term. We investigate both classical convex regularizations such as ℓ1 or weighted ℓ1 and non-convex regularization terms such as log penalty, Minimax Concave Penalty (MCP) or ℓp pseudo norm with p < 1. Two algorithms are proposed, first an accelerated proximal approach for solving the convex problems, second a reweighted ℓ1 scheme to address the non-convex regularizations. We conduct intensive experiments on nine datasets from Letor 3.0 and Letor 4.0 corpora. Numerical results show that the use of non-convex regularizations we propose leads to more sparsity in the resulting models while prediction performance is preserved. The number of features is decreased by up to a factor of six compared to the ℓ1 regularization. In addition, the software is publicly available on the web. 1 2 #{(ds,dt)∈D|dt≻xds and dt≻yds} #{(ds,dt)∈D}
Journal Européen des Systèmes Automatisés, 2005
Résumé/Abstract The paper investigates the application of black box modelling to the prediction o... more Résumé/Abstract The paper investigates the application of black box modelling to the prediction of the daily maxima of ground-ozone level. The main interest of these modelling approaches is their genericity as they are solely based on the available data provided by the Associations of air quality monitoring and they can be transposed from a geographical area to another one. The paper realises a comparative study of four statistical learning approaches, the decisions trees (, the neural networks, the least-angle regression and the ...
2013 Humaine Association Conference on Affective Computing and Intelligent Interaction, 2013
ABSTRACT The objective of the paper is to investigate the effect of subject's emotional s... more ABSTRACT The objective of the paper is to investigate the effect of subject's emotional states on Brain Computer Interface (BCI) performance. Two psycho-physiological experiments are designed and implemented. The first one induces subjects' emotion using video clips first, then involves subjects' in SSVEP task. The second one induces subjects' emotions and SSVEP simultaneously by flickering IAPS pictures in four directions. used to recognize the performed BCI tasks. Based on the performances of learned classifiers, we analyzed the influence of emotion using two statistical tests. The McNamara's test serves to assess if emotion has any influences on mental task performing while Wilcox on signed-rank test analyses if emotion has a positive or detrimental effect on ability to achieve a SSVEP task. Obtained results suggest influence of emotional states: the positive and neutral emotions influence BCI performance similarly, while the negative emotion tends to deteriorate classification accuracy.