Showing papers by "Mines ParisTech published in 2007"

Comparison of Pansharpening Algorithms: Outcome of the 2006 GRS-S Data-Fusion Contest

Abstract: In January 2006, the Data Fusion Committee of the IEEE Geoscience and Remote Sensing Society launched a public contest for pansharpening algorithms, which aimed to identify the ones that perform best. Seven research groups worldwide participated in the contest, testing eight algorithms following different philosophies [component substitution, multiresolution analysis (MRA), detail injection, etc.]. Several complete data sets from two different sensors, namely, QuickBird and simulated Pleiades, were delivered to all participants. The fusion results were collected and evaluated, both visually and objectively. Quantitative results of pansharpening were possible owing to the availability of reference originals obtained either by simulating the data collected from the satellite sensor by means of higher resolution data from an airborne platform, in the case of the Pleiades data, or by first degrading all the available data to a coarser resolution and saving the original as the reference, in the case of the QuickBird data. The evaluation results were presented during the special session on data fusion at the 2006 international geoscience and remote sensing symposium in Denver, and these are discussed in further detail in this paper. Two algorithms outperform all the others, the visual analysis being confirmed by the quantitative evaluation. These two methods share the same philosophy: they basically rely on MRA and employ adaptive models for the injection of high-pass details.

More efficiency in multiple kernel learning

Abstract: An efficient and general multiple kernel learning (MKL) algorithm has been recently proposed by Sonnenburg et al. (2006). This approach has opened new perspectives since it makes the MKL approach tractable for large-scale problems, by iteratively using existing support vector machine code. However, it turns out that this iterative algorithm needs several iterations before converging towards a reasonable solution. In this paper, we address the MKL problem through an adaptive 2-norm regularization formulation. Weights on each kernel matrix are included in the standard SVM empirical risk minimization problem with a l1 constraint to encourage sparsity. We propose an algorithm for solving this problem and provide an new insight on MKL algorithms based on block 1-norm regularization by showing that the two approaches are equivalent. Experimental results show that the resulting algorithm converges rapidly and its efficiency compares favorably to other MKL algorithms.

Coagulation and flocculation of dye-containing solutions using a biopolymer (Chitosan)

Abstract: Chitosan, dissolved in acetic acid, was used for the coagulation–flocculation of an anionic dye (Reactive Black 5). In acidic solutions protonated amine groups of chitosan attract dye sulfonic groups. Increasing chitosan dosage increases dye removal up to a concentration resulting in complete neutralization of anionic charges; above, the excess of cationic charges leads to suspension re-stabilization. Process efficiency increases with decreasing the initial pH of dye solution: the molar ratio between dye molecules and amine groups ([ n ]) respects the stoichiometry between sulfonic functions and protonated amine groups at initial pH 5; at initial pH 3 a possible dye aggregation phenomenon results in higher molar ratio [ n ]. The coefficient [ n ] depends on both the pH and the molecular weight of chitosan. The main mechanism for dye coagulation with chitosan sounds to be charge neutralization at acidic pH.

Non‐parametric probabilistic forecasts of wind power: required properties and evaluation

Abstract: Predictions of wind power production for horizons up to 48-72 h ahead comprise a highly valuable input to the methods for the daily management or trading of wind generation. Today, users of wind power predictions are not only provided with point predictions, which are estimates of the conditional expectation of the wind generation for each look-ahead time, but also with uncertainty estimates given by probabilistic forecasts. In order to avoid assumptions on the shape of predictive distributions, these probabilistic predictions are produced from non-parametric methods, and then take the form of a single or a set of quantile forecasts. The required and desirable properties of such probabilistic forecasts are defined and a framework for their evaluation is proposed. This framework is applied for evaluating the quality of two statistical methods producing full predictive distributions from point predictions of wind power. These distributions are defined by a number of quantile forecasts with nominal proportions spanning the unit interval. The relevance and interest of the introduced evaluation framework are discussed.

Classification of microarray data using gene networks

Abstract: Microarrays have become extremely useful for analysing genetic phenomena, but establishing a relation between microarray analysis results (typically a list of genes) and their biological significance is often difficult. Currently, the standard approach is to map a posteriori the results onto gene networks in order to elucidate the functions perturbed at the level of pathways. However, integrating a priori knowledge of the gene networks could help in the statistical analysis of gene expression data and in their biological interpretation. We propose a method to integrate a priori the knowledge of a gene network in the analysis of gene expression data. The approach is based on the spectral decomposition of gene expression profiles with respect to the eigenfunctions of the graph, resulting in an attenuation of the high-frequency components of the expression profiles with respect to the topology of the graph. We show how to derive unsupervised and supervised classification algorithms of expression profiles, resulting in classifiers with biological relevance. We illustrate the method with the analysis of a set of expression profiles from irradiated and non-irradiated yeast strains. Including a priori knowledge of a gene network for the analysis of gene expression data leads to good classification performance and improved interpretability of the results.

A Kernel for Time Series Based on Global Alignments

Abstract: We propose in this paper a new family of kernels to handle time series, notably speech data, within the framework of kernel methods which includes popular algorithms such as the support vector machine. These kernels elaborate on the well known dynamic time warping (DTW) family of distances by considering the same set of elementary operations, namely substitutions and repetitions of tokens, to map a sequence onto another. Associating to each of these operations a given score, DTW algorithms use dynamic programming techniques to compute an optimal sequence of operations with high overall score, in this paper we consider instead the score spanned by all possible alignments, take a smoothed version of their maximum and derive a kernel out of this formulation. We prove that this kernel is positive definite under favorable conditions and show how it can be tuned effectively for practical applications as we report encouraging results on a speech recognition task.

An exact algorithm for team orienteering problems

Abstract: Optimising routing of vehicles constitutes a major logistic stake in many industrial contexts. We are interested here in the optimal resolution of special cases of vehicle routing problems, known as team orienteering problems. In these problems, vehicles are guided by a reward that can be collected from customers, while the length of routes is limited. The main difference with classical vehicle routing problems is that not all customers have to be visited. The solution method we propose here is based on a Branch & Price algorithm. It is, as far as we know, the first exact method proposed for such problems, except for a preliminary work from Gueguen (Methodes de resolution exacte pour problemes de tournees de vehicules. These de doctorat, ecole Centrale Paris 1999) and a work from Butt and Ryan (Comput Oper Res 26(4):427–441 1999). It permits to solve instances with up to 100 customers.

Scanning Electron Microscopy for Quantitative Small and Large Deformation Measurements Part II: Experimental Validation for Magnifications from 200 to 10,000

Abstract: A combination of drift distortion removal and spatial distortion removal are performed to correct Scanning Electron Microscope (SEM) images at both ×200 and ×10,000 magnification. Using multiple, time-spaced images and in-plane rigid body motions to extract the relative displacement field throughout the imaging process, results from numerical simulations clearly demonstrate that the correction procedures successfully remove both drift and spatial distortions with errors on the order of ±0.02 pixels. A series of 2D translation and tensile loading experiments are performed in an SEM for magnifications at ×200 and ×10,000, where both the drift and spatial distortion removal methods described above are applied to correct the digital images and improve the accuracy of measurements obtained using 2D-DIC. Results from translation and loading experiments indicate that (a) the fully corrected displacement components have nearly random variability with standard deviation of 0.02 pixels (≈25 nm at ×200 and ≈0.5 nm at ×10,000) in each displacement component and (b) the measured strain fields are unbiased and in excellent agreement with expected results, with a spatial resolution of 43 pixels (≈54 μm at ×200 and ≈1.1 μm at ×10,000) and a standard deviation on the order of 6 × 10−5 for each component.

Geochemistry of Neoproterozoic shales of the Rabanpalli Formation, Bhima Basin, Northern Karnataka, southern India: implications for provenance and paleoredox conditions

Abstract: The Rabanpalli Formation exhibits two types of shales, viz. grey and calcareous shales. These shales (grey and calcareous) have been analysed for major, trace, and rare earth elements to fi nd out their source rocks characteristics and paleo-oxygenation conditions. The grey shales have higher concentration of SiO2, Al2O3, Fe2O3, K2O, Zr, Th, U, V, Cr, La, Ce, and Y than calcareous shales, whereas calcareous shales are enriched in CaO, Mn, Sr, Ba, Cu, and Zn, which indicate that the carbonate phase minerals are higher in calcareous shales. The positive correlation of K2O with other elements, and abundance of Al2O3, Ba, Th, and Rb suggest that these elements are primarily controlled by the dominant clay minerals. La/Sc, Th/Sc, Th/Co, Th/Cr, and Cr/Th ratios of shales were compared with those of sediments derived from felsic and basic rocks (fi ne fraction), upper continental crust (UCC) and post-Archean Australian average shale (PAAS) ratios, which reveal that these ratios are within the range of felsic rocks. The La/Sc vs. Th/Co plot also suggests the felsic nature of the source rocks. The shales show slightly light rare earth element (LREE) enriched and fl at heavy rare earth element (HREE) patterns with negative Eu anomaly, and are similar to the granitoids from Dharwar Craton, which suggest that the Archean Dharwar Craton contributed the sediments to the Bhima basin. The geochemical parameters such as U, authigenic U, U/Th, V/Cr, Ni/Co, and Cu/Zn ratios indicate that these shales were deposited under oxic environment.

Probabilistic Short-term Wind Power Forecasting for the Optimal Management of Wind Generation

Abstract: Wind power forecasting tools have been developed for some time. The majority of such tools usually provides single-valued (spot) predictions. Such predictions limits the use of tools for decision-making under uncertainty. In this paper we propose a method for producing the complete predictive probability density function (PDF). The method is based on kernel density estimation techniques. The preliminary results show that this method levels with state of the art one while being fast and producing the complete PDF. The results were obtained through real data from three French wind farms.

OrthoMaM: A database of orthologous genomic markers for placental mammal phylogenetics

Abstract: Molecular sequence data have become the standard in modern day phylogenetics. In particular, several long-standing questions of mammalian evolutionary history have been recently resolved thanks to the use of molecular characters. Yet, most studies have focused on only a handful of standard markers. The availability of an ever increasing number of whole genome sequences is a golden mine for modern systematics. Genomic data now provide the opportunity to select new markers that are potentially relevant for further resolving branches of the mammalian phylogenetic tree at various taxonomic levels. The EnsEMBL database was used to determine a set of orthologous genes from 12 available complete mammalian genomes. As targets for possible amplification and sequencing in additional taxa, more than 3,000 exons of length > 400 bp have been selected, among which 118, 368, 608, and 674 are respectively retrieved for 12, 11, 10, and 9 species. A bioinformatic pipeline has been developed to provide evolutionary descriptors for these candidate markers in order to assess their potential phylogenetic utility. The resulting OrthoMaM (Orthologous Mammalian Markers) database can be queried and alignments can be downloaded through a dedicated web interface http://kimura.univ-montp2.fr/orthomam . The importance of marker choice in phylogenetic studies has long been stressed. Our database centered on complete genome information now makes possible to select promising markers to a given phylogenetic question or a systematic framework by querying a number of evolutionary descriptors. The usefulness of the database is illustrated with two biological examples. First, two potentially useful markers were identified for rodent systematics based on relevant evolutionary parameters and sequenced in additional species. Second, a complete, gapless 94 kb supermatrix of 118 orthologous exons was assembled for 12 mammals. Phylogenetic analyses using probabilistic methods unambiguously supported the new placental phylogeny by retrieving the monophyly of Glires, Euarchontoglires, Laurasiatheria, and Boreoeutheria. Muroid rodents thus do not represent a basal placental lineage as it was mistakenly reasserted in some recent phylogenomic analyses based on fewer taxa. We expect the OrthoMaM database to be useful for further resolving the phylogenetic tree of placental mammals and for better understanding the evolutionary dynamics of their genomes, i.e., the forces that shaped coding sequences in terms of selective constraints.

Left-invariant extended Kalman filter and attitude estimation

Abstract: We consider a left-invariant dynamics on a Lie group. One way to define driving and observation noises is to make them preserve the symmetries. We propose a left- invariant (i.e, intrinsic and thus symmetry-preserving) extended Kalman filter such that the left-invariant estimation error obeys a stochastic differential equation independent of the system trajectory. The theory is illustrated by an attitude estimation example.

A new pairwise kernel for biological network inference with support vector machines

Abstract: Much recent work in bioinformatics has focused on the inference of various types of biological networks, representing gene regulation, metabolic processes, protein-protein interactions, etc. A common setting involves inferring network edges in a supervised fashion from a set of high-confidence edges, possibly characterized by multiple, heterogeneous data sets (protein sequence, gene expression, etc.). Here, we distinguish between two modes of inference in this setting: direct inference based upon similarities between nodes joined by an edge, and indirect inference based upon similarities between one pair of nodes and another pair of nodes. We propose a supervised approach for the direct case by translating it into a distance metric learning problem. A relaxation of the resulting convex optimization problem leads to the support vector machine (SVM) algorithm with a particular kernel for pairs, which we call the metric learning pairwise kernel. This new kernel for pairs can easily be used by most SVM implementations to solve problems of supervised classification and inference of pairwise relationships from heterogeneous data. We demonstrate, using several real biological networks and genomic datasets, that this approach often improves upon the state-of-the-art SVM for indirect inference with another pairwise kernel, and that the combination of both kernels always improves upon each individual kernel. The metric learning pairwise kernel is a new formulation to infer pairwise relationships with SVM, which provides state-of-the-art results for the inference of several biological networks from heterogeneous genomic data.

Supervised reconstruction of biological networks with local models

Abstract: Motivation: Inference and reconstruction of biological networks from heterogeneous data is currently an active research subject with several important applications in systems biology. The problem has been attacked from many different points of view with varying degrees of success. In particular, predicting new edges with a reasonable false discovery rate is highly demanded for practical applications, but remains extremely challenging due to the sparsity of the networks of interest. Results: While most previous approaches based on the partial knowledge of the network to be inferred build global models to predict new edges over the network, we introduce here a novel method which predicts whether there is an edge from a newly added vertex to each of the vertices of a known network using local models. This involves learning individually a certain subnetwork associated with each vertex of the known network, then using the discovered classification rule associated with only that vertex to predict the edge to the new vertex. Excellent experimental results are shown in the case of metabolic and protein–protein interaction network reconstruction from a variety of genomic data. Availability: An implementation of the proposed algorithm is available upon request from the authors. Contact: Jean-Philippe.Vert@ensmp.fr

Impact of coating development on the hydraulic and transport properties in argillaceous limestone fracture

Abstract: [1] Results are reported for an acidic water flow-through experiment conducted in a fractured argillaceous limestone sample (73% carbonates) The change in fracture geometry and related parameters is reported for six data sets obtained from synchrotron X-ray microtomography experiments High-resolution three-dimensional images of the sample allowed quantification of the changes in fracture morphology at a spatial resolution of 6 μm Mineral mass loss and permeability changes in the sample were also determined Several physico-chemical phenomena were identified during the experiment Initial smooth fracture surfaces evolved rapidly toward rough surfaces with uneven clay coverage due to the preferential dissolution of carbonate minerals compared to clay minerals whose dissolution rate is about 106 slower A microporous clay coating progressively developed at the fluid-rock interface during heterogeneous dissolution of the fracture, while the global dissolution rate of the fracture walls exponentially decreased The increase in surface roughness and the presumed reorganization of clays caused a progressive reduction in permeability During the last flow-through stage, a large decrease in sample permeability was attributed to the large removal of clay particles; this process was responsible for a dramatic collapse of the fracture walls near the sample inlet and led to the development of preferential flow pathways The development of the clay coating also acted as a barrier to flow and mass transfer between calcite grains and bulk solution and affected transport processes within the fracture

Stress overshoots of organoclay nanocomposites in transient shear flow

Abstract: Forward and reverse stress growth experiments have been conducted on polypropylene/organoclay nanocomposites containing the same clay loading but characterized by different microstructures. Stress overshoots have been observed for the initial start-up experiments and for the following reverse start-up experiments after a certain rest time. The amplitude of these overshoots increased with the applied shear rate and rest time, but the overshoots occurred at the same strain of about 1.7. The overshoots are related to the structure of the nanocomposites, in particular the magnitude of the overshoots increased with the degree of the clay exfoliation in the matrix. Two models, initially developed for colloidal suspensions and fiber suspensions, have been used to describe the observed phenomena. The overshoots are fairly well predicted by the first (structure network) model and explained by the competing effects of the structure breakdown under flow and reorganization during rest time. However, the model predicts that the shear stress following the overshoot decreases and reaches steady-state too rapidly. The second model developed for ellipsoid suspensions describes quite well the stress overshoots for the initial forward flow, but no effect of rest time is predicted. A modified version has been proposed by adding a molecular diffusivity contribution in the Folgar-Tucker equation. The effect of the particle disorientation is qualitatively predicted, but the kinetics is too slow compared to that deduced from experiments.

Flatness-Based Vehicle Steering Control Strategy With SDRE Feedback Gains Tuned Via a Sensitivity Approach

Abstract: This paper presents a feedback steering control strategy for a vehicle in an automatic driving context. Two main contributions in terms of control are highlighted. On the one hand, the automatic reference trajectories generation from geometric path constraints (obstacles). Thanks to the flatness property of the considered model, the longitudinal velocity will be controlled around a quasi-constant value while lateral and yaw dynamics targets will allow to avoid obstacles. On the other hand, a sensitivity-based methodology will be presented to choose the best possible gains parameterization in a state Riccati dependent equation (SDRE) feedback controller. Both direct and adjoint sensitivity methods are used, together with a dynamic inversion of the system, in order to optimize the performances of the controller. Obstacle avoiding simulation results will be validated and compared with other nonlinear optimal feedback controllers, from a realistic industrial simulator environment for vehicle dynamics