Entity

Time filter

Source Type

Saint-Calais, France

Kalakech M.,Heilongjiang University | Kalakech M.,Lille University of Science and Technology | Biela P.,Heilongjiang University | Biela P.,Lille University of Science and Technology | And 2 more authors.
Pattern Recognition Letters | Year: 2011

Recent feature selection scores using pairwise constraints (must-link and cannot-link) have shown better performances than the unsupervised methods and comparable to the supervised ones. However, these scores use only the pairwise constraints and ignore the available information brought by the unlabeled data. Moreover, these constraint scores strongly depend on the given must-link and cannot-link subsets built by the user. In this paper, we address these problems and propose a new semi-supervised constraint score that uses both pairwise constraints and local properties of the unlabeled data. Experiments using Kendall's coefficient and accuracy rates, show that this new score is less sensitive to the given constraints than the previous scores while providing similar performances.© 2010 Elsevier B.V. All rights reserved. Source


Hijazi H.,Universite Ibn Tofail | Bazzi O.,Universite Ibn Tofail | Bigand A.,LISIC
Proceedings of the 2011 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2011 | Year: 2011

Linear discriminant analysis (LDA) is a simple but widely used algorithm in pattern recognition. However it has some shortcomings in that it is sensitive to outliers and limited to linearly separable cases. To solve this problem a new version of nonlinear discriminant algorithm is proposed. This new version, SC-LLE, uses LDA combined with LLE method to take into account non-linearly separable cases. There have been other attempts to solve the problems of LDA including methods using kernels (KPCA). However, we investigate this new concept applied to nonlinear data projection that seems to be promising. Applications on 3D data show the interest of this concept. Source


Kalakech M.,Lebanese University | Biela P.,Heilongjiang University | Hamad D.,LISIC | Macaire L.,Lille University of Science and Technology
Neural Processing Letters | Year: 2013

Semi-supervised context characterized by the presence of a few pairs of constraints between learning samples is abundant in many real applications. Analysing these instance constraints by recent spectral scores has shown good performances for semi-supervised feature selection. The performance evaluation of these scores is generally based on classification accuracy and is performed in a ground truth context. However, this supervised context used by the evaluation step is inconsistent with the semi-supervised context in which the feature selection operates. In this paper, we propose a semi-supervised performance evaluation procedure, so that both feature selection and clustering steps take into account the constraints given by the user. In this way, the selection and the evaluation steps are performed in the same context which is close to real life applications. Extensive experiments on benchmark datasets are carried out in the last section. These experiments are performed using a supervised classical evaluation and the semi-supervised proposed one. They demonstrate the effectiveness of feature selection based on constraint analysis that uses both pairwise constraints and the information brought by the unlabeled data. © 2013 Springer Science+Business Media New York. Source


Hijazi H.,Universite Ibn Tofail | Bazzi O.,Universite Ibn Tofail | Bigand A.,LISIC
2013 3rd International Conference on Communications and Information Technology, ICCIT 2013 | Year: 2013

In a previous paper a new version of nonlinear dimensionality reduction algorithm was proposed, the SC-LLE approach. This approach combines a supervised method, linear discriminant analysis (LDA, a simple but widely used algorithm in pattern recognition) with an unsupervised method, local linear embedding (LLE, manifold learning). SC-LLE method can generalize any linear classifier (like LDA) to nonlinear by transforming data into some low-dimensional feature space. This new concept (SC-LLE) applied to nonlinear data projection seems to be promising, and we show in this new paper that semi-supervised learning (SSL) is another interesting property of SC-LLE. Applications on 3D data show the interest of this method. © 2013 IEEE. Source


Bourgois L.,Supelec | Roussel G.,LISIC | Benjelloun M.,LISIC
Neural Networks | Year: 2013

This paper deals with the design methodology of an Inverse Neural Network (INN) model. The basic idea is to carry out a semi-physical model gathering two types of information: the a priori knowledge of the deterministic rules which govern the studied system and the observation of the actual conduct of this system obtained from experimental data. This hybrid model is elaborated by being inspired by the mechanisms of a neuromimetic network whose structure is constrained by the discrete reverse-time state-space equations. In order to validate the approach, some tests are performed on two dynamic models. The first suggested model is a dynamic system characterized by an unspecified r-order Ordinary Differential Equation (ODE). The second one concerns in particular the mass balance equation for a dispersion phenomenon governed by a Partial Differential Equation (PDE) discretized on a basic mesh. The performances are numerically analyzed in terms of generalization, regularization and training effort. © 2012 Elsevier Ltd. Source

Discover hidden collaborations