Browsing by Author "Fouzi Douak"

Now showing 1 - 9 of 9
  • No Thumbnail Available
    Item
    A Modular Support Vector Machine for Active Learning of Urban Remote Sensing Images Classification in Algeria
    (Springer - Journal of the Indian Society of Remote Sensing, 2018) Fouzi Douak
    In this work, we present a new strategy of active learning, based on a modular version of support vector machine (MSVM) applied to urban remote sensing images in Algeria. In general, the training set is highly imbalanced, which gives more complex models; this difficulty is solved by dividing the problem at hand into a set of sub-problems, where each sub-model could be simpler to solve. The support vector machine is introduced to solve the problem of classification based on image remote sensing data related to atmospheric conditions and illumination reflectance. The aim of the proposed method is to improve the accuracy in order to understand the correlated elements of urban structures (the site, the built, the parcels, the network, the space), to generate the final classification result. In particular, we propose a new method based on the modular support vector machine (MSVM) adopted to active learning method, using three different clustering methods (i) k-means, (ii) fuzzy c-means (FCM), and (iii) Gustafson–Kessel (GKclust). Experimental results obtained on two QuickBird multispectral images of Se´tif and Batna cities in the eastern of Algeria confirm the capabilities of the proposed methods based on the ensemble of model trained with different task decomposition compared to a traditional model using active learning. This method improves each class presents a main register in urban structure tissues.
  • No Thumbnail Available
    Item
    A two-stage regression approach for spectroscopic quantitative analysis
    (Elsevier - Chemometrics and Intelligent Laboratory Systems, 2011) Fouzi Douak
    In this paper, we propose a two-stage regression approach,which is based on the residual correction concept. Its underlying idea is to correct any given regressor by analyzing and modeling its residual errors in the input space. Wereport and discuss results of experiments conducted on three different datasets in infrared spectroscopy and designed in such away to test the proposed approach by: 1) varying the kind of adopted regressionmethod used to approximate the chemical parameter of interest. Partial least squares regression (PLSR), support vector machines (SVM) and radial basis function neural network (RBF) methods are considered; 2) adopting or not a feature selection strategy to reduce the dimension of the space where to perform the regression task. A comparative studywith another approachwhich exploits differently estimation errors, namely adaptive boosting for regression (AdaBoost.R), is also included. The obtained results point out that the residual-based correction approach (RBC) can improve the accuracy of the estimation process. Not all the improvements are statistically significant but, at the same time, no case of accuracy decrease has been observed.
  • No Thumbnail Available
    Item
    Active learning for spectroscopic data regression
    (Wiley - J. Chemometrics, 2012) Fouzi Douak
    In this work, we introduce an active learning approach for the estimation of chemical concentrations from spectroscopic data. Its main objective is to opportunely collect training samples in such a way as to minimize the error of the regression process while minimizing the number of training samples used, and thus to reduce the costs related to training sample collection. In particular, we propose two different active learning strategies developed for regression approaches based on partial least squares regression, ridge regression, kernel ridge regression, and support vector regression. The first strategy uses a pool of regressors in order to select the samples with the greatest disagreements among the different regressors of the pool, while the second one is based on adding samples that are distant from the current training samples in the feature space. For support vector regression, a specific strategy based on the selection of the samples distant from the support vectors is proposed. Experimental results on three different real data sets are reported and discussed.
  • No Thumbnail Available
    Item
    Color imagecompressionalgorithmbasedontheDCTtransformcombinedto an adaptiveblockscanning
    (Elsevier - Int. J.Electron.Commun.(AEU¨ ), 2011) Fouzi Douak
    This paper considers the design of a lossy image compression algorithm dedicated to color still images. After a preprocessing step (mean removing and RGB to YCbCr transformation), the DCT transform is applied and followed by an iterative phase (using the bisection method) including the thresholding, the quantization, dequantization, the inverse DCT, YCbCr to RGB transform and the mean recovering. This is done in order to guarantee that a desired quality (fixed in advance using the well known PSNR metric) is checked. For the aim to obtain the best possible compression ratio CR, the next step is the application of a proposed adaptive scanning providing, for each (n, n) DCT block a corresponding (nxn) vector containing the maximum possible run of zeros at its end. The last step is the application of a modified systematic lossless encoder. The efficiency of the proposed scheme is demonstrated by results, especially, when faced to the method presented in the recently published paper based on the block truncation coding using pattern fitting principle.
  • No Thumbnail Available
    Item
    Genetic robust kernel sample selection for chemometric data analysis
    (Wiley Journal of Chemometrics, 2021) Fouzi Douak
    In this work, we propose a new algorithm to improve existing techniques used in the field of spectroscopic data regression analysis. In particular, it combines the power of nonlinear kernel regressors (kernel ridge regression [KRR], kernel principal component regression [KPCR], and Gaussian process regression [GPR]) with an optimization based on nondominated sorting multi-objective genetic algorithm (NSGAII) to filter the residual outliers in the prediction space and leverage points in the features space. The proposed algorithm, contrary to most existing robust algorithms, simultaneously optimizes many complementary objectives for an automatic adaptation and thus a better outliers detection. It is well known that the elimination of outliers greatly improves the regression model. It is thus the aim of this work to develop a new robust regression algorithm. It has been applied on five different datasets, and the results are compared to both classical nonlinear regression methods and the commonly used robust regression methods robust continuum regression (RCR), partial robust M-regression (PRM), robust principal component regression (RPCR), robust PLSR (RSIMPLS), and locally weighted regression (LWR). They show that the proposed algorithm outperforms the classical nonlinear regression methods and is a promising competitor to the robust methods outperforming most of them. Even though the results obtained are only from five datasets, this algorithm can be considered an interesting contribution for improving data analysis in the field of chemometrics.
  • No Thumbnail Available
    Item
    Kernel ridge regression with active learning for wind speed prediction
    (Elsevier - Applied Energy, 2013) Fouzi Douak
    This paper introduces the active learning approach for wind speed prediction. The main objective of active learning is to opportunely collect training samples in such a way as to minimize the error of the prediction process while minimizing the number of training samples used, and thus to reduce the costs related to the training sample collection. In particular, we propose three different active learning strategies, developed for kernel ridge regression (KRR). The first strategy uses a pool of regressors in order to select the samples with the greater disagreements between the different regressors, while the second one relies on the idea to add samples that are distant from the available training samples, and the last strategy is based on the selection of samples which exhibit a high expected prediction error. A thorough experimental study is presented. It is based on ten different wind speed measurement stations distributed over the vast Algerian territory. Promising results are reported, showing that a smart collection of training samples can be of benefit for wind speed prediction problems.
  • No Thumbnail Available
    Item
    Role of non-uniform channel doping in improving the nanoscale JL DG MOSFET reliability against the self-heating effects
    (Elsevier - Superlattices and Microstructures, 2017) Fouzi Douak
    In this paper, a new hybrid approach by combining numerical investigation and Support Vector Machines (SVMs) classifier is proposed to study the thermoelectric performance of nanoscale Double Gate Junctionless DG JL MOSFET. In this context, a new Figure of Merit (FoM) parameter which combines both electrical and reliability characteristics is proposed. Moreover, the impact of Gaussian channel doping profile (GCD) in enhancing the DG JL MOSFET reliability against the self-heating effects (SHEs) is presented. The proposed design thermal stability and electrical characteristics are investigated and compared with those of the conventional structure in order to reveal the device performance including SHEs. It is found that the amended channel doping has a profound implication in improving both the device electrical performance and the reliability against the undesired self-heating and short channel effects (SCEs). Furthermore, the transistor thermal behavior analysis involves classification of the device performance by taking into account the device reliability. For this purpose, SVMs are adopted for supervised classification in order to identify the most favorable design configurations associated with suppressed SHEs and improved electrical performance. We find that the proposed design methodology has succeeded in selecting the better designs that offer superior reliability against the SHEs. The obtained results suggest the possibility for bridging the gap between high electrical performances with better immunity to the SHEs.
  • No Thumbnail Available
    Item
    Support Vector Regression in Spectrophotometry: An Experimental Study
    (Taylor and Francis - Critical Reviews in Analytical Chemistry, 2012) Fouzi Douak
    In this work, we present a detailed experimental assessment of an interesting regression approach based on support vector machines (SVMs), a technique relatively recently introduced in the literature. The experimental framework reports a thorough investigation of the performance of SVMs from different viewpoints, including: (i) the influence of the kernel type in the SVM regression task; (ii) the sensitivity to the number of input variables (spectra dimension); (iii) the sensitivity to the available number of training samples; and (iv) the overall stability. The obtained results are compared with those yielded by the radial basis function (RBF) and the multilayer perceptron (MLP) neural networks as well as the traditional multiple linear regression (MLR) method on two different spectrophotometric datasets.
  • No Thumbnail Available
    Item
    Towards a NIR Spectroscopy ensemble learning technique competing with the standard ASTM-CFR: An optimal boosting and bagging extreme learning machine algorithms for gasoline octane number prediction
    (Optik -International Journal for Light and Electron Optics, 2022) Fouzi Douak
    The determination of the octane number plays a major role in quantifying the quality of gasoline. The standard method, the international standard ASTM-CFR internal combustion engine, used for this purpose suffers from its high cost and time. Many algorithms have been developed to address the limitations of this method, taking advantage of infrared spectroscopy, which provides easily measurable parameters that can be used to predict the octane number. This paper proposes two methods to compete with three existing algorithms (ELM, IELM, and SaDE-ELM) and also aims at achieving high accuracy results compared to the engine-based measurement method. The proposed methods used ensemble learning strategy combined with ELM instead of single ELM learner, used in the aforementioned algorithms, to achieve better predictions of the octane number. The findings indicate our algorithms outperformed the existing algorithms and the predictions are very close to that of the standard method. This can be considered an important achievement in the field of octane number prediction that can eventually replace the standard method. Also, contrary to the common belief that Boosting algorithms are superior to Bagging algorithms, in this paper, we demonstrated that our Bagging algorithm performed almost identically compared with the Boosting algorithm. Despite these promising results, further research should be undertaken to investigate the Hughes effect phenomenon that the gasoline data set, used in this work, suffer from.