2007 IEEE Workshop on Machine Learning for Signal Processing

27-29 Aug. 2007

Filter Results

Displaying Results 1 - 25 of 83
  • [Advertisement]

    Publication Year: 2007, Page(s): nil1
    Request permission for commercial reuse | PDF file iconPDF (289 KB)
    Freely Available from IEEE
  • Copyright page

    Publication Year: 2007, Page(s): nil1
    Request permission for commercial reuse | PDF file iconPDF (289 KB)
    Freely Available from IEEE
  • MLSP 2007 Technical Committee

    Publication Year: 2007, Page(s): nil2
    Request permission for commercial reuse | PDF file iconPDF (34 KB)
    Freely Available from IEEE
  • Copyright page

    Publication Year: 2007, Page(s): nil3
    Request permission for commercial reuse | PDF file iconPDF (158 KB)
    Freely Available from IEEE
  • [Front cover]

    Publication Year: 2007, Page(s): C1
    Request permission for commercial reuse | PDF file iconPDF (94 KB)
    Freely Available from IEEE
  • Preface

    Publication Year: 2007, Page(s): II
    Request permission for commercial reuse | PDF file iconPDF (40 KB) | HTML iconHTML
    Freely Available from IEEE
  • [Advertisement]

    Publication Year: 2007, Page(s): nil4
    Request permission for commercial reuse | PDF file iconPDF (99 KB)
    Freely Available from IEEE
  • Proceedings of the 2007 IEEE Signal Processing Society Workshop

    Publication Year: 2007, Page(s): nil5
    Request permission for commercial reuse | PDF file iconPDF (170 KB)
    Freely Available from IEEE
  • Table of contents

    Publication Year: 2007, Page(s):IV - IX
    Request permission for commercial reuse | PDF file iconPDF (142 KB)
    Freely Available from IEEE
  • On Feature Selection for Genomic Signal Processing and Data Mining

    Publication Year: 2007, Page(s):1 - 20
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (6174 KB) | HTML iconHTML

    An effective data mining system lies in the representation of pattern vectors. The most vital information to be represented is the characteristics embedded in the raw data most essential for the intended applications. In order to extract a useful high-level representation, it is desirable that a representation can provide concise, invariant, and/or intelligible information on input patterns. The c... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Fast Network Component Analysis for Gene Regulation Networks

    Publication Year: 2007, Page(s):21 - 26
    Cited by:  Papers (3)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (1912 KB) | HTML iconHTML

    New advancement in microarray technologies has made it possible to reconstruct gene regulation networks from mass gene expression data measured by microarray. Typically, gene regulation networks are sparse networks. This sparse topology knowledge can be exploited to develop algorithms for network reconstruction. In this direction, a method called network component analysis (NCA) has been developed... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Discriminant Subspaces of Some High Dimensional Pattern Classification Problems

    Publication Year: 2007, Page(s):27 - 32
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (2272 KB) | HTML iconHTML

    In this paper, we report on an empirical study of several high dimensional classification problems and show that much of the discriminant information may lie in low dimensional subspaces. Feature subset selection is achieved either by forward selection or backward elimination from the full feature space with support vector machines (SVMs) as base classifiers. These "wrapper" methods are compared w... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multiplicative Updates for the Lasso

    Publication Year: 2007, Page(s):33 - 38
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (467 KB) | HTML iconHTML

    Multiplicative updates have proven useful for non-negativity constrained optimization. Presently, we demonstrate how multiplicative updates also can be used for unconstrained optimization. This is for instance useful when estimating the least absolute shrinkage and selection operator (LASSO) i.e. least squares minimization with L1-norm regularization, since the multiplicative updates (M... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Sensitivity Analysis of Boosting PSI-Blast with Case Study on Subcellular Localization

    Publication Year: 2007, Page(s):39 - 44
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (3503 KB) | HTML iconHTML

    This paper studies the sensitivity of PSI-BLAST with respect to the 'h' parameter. Observing that the standard PSI-BLAST is sensitive to parameter 'h' in the high-value region, we propose a new technique, called boosting PSI- BLAST, to reduce the sensitivity. By constraining 'h' to a small value first so as to reduce the chance of early corruption and then relaxing it gradually to increase diverge... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • MLSP 2007 Data Analysis Competition: Frequency-Domain Blind Source Separation for Convolutive Mixtures of Speech/Audio Signals

    Publication Year: 2007, Page(s):45 - 50
    Cited by:  Papers (7)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (1979 KB) | HTML iconHTML

    This paper describes the frequency-domain approach to the blind source separation of speech/audio signals that are convolutively mixed in a real room environment. With the application of short- time Fourier transforms, convolutive mixtures in the time domain can be approximated as multiple instantaneous mixtures in the frequency domain. We employ complex-valued independent component analysis (ICA)... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • MLSP 2007 Data Analysis Competition: Two-Stage Blind Source Separation Combining SIMO-Model-Based ICA and Binary Masking

    Publication Year: 2007, Page(s):51 - 56
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (3469 KB) | HTML iconHTML

    This paper reviews a real-time two-stage blind source separation (BSS) method for convolutive mixtures of speech, in which a single-input multiple-output (SIMO)-model-based independent component analysis (ICA) and a SIMO-model-based binary masking are combined. SIMO-model-based ICA can separate the mixed signals, not into monaural source signals but into SIMO-model-based signals from independent s... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Local Subspace Classifiers: Linear and Nonlinear Approaches

    Publication Year: 2007, Page(s):57 - 62
    Cited by:  Papers (6)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (648 KB) | HTML iconHTML

    The K-local hyperplane distance nearest neighbor (HKNN) algorithm is a local classification method which builds nonlinear decision surfaces directly in the original sample space by using local linear manifolds. Although the HKNN method has been successfully applied in several classification tasks, it is not possible to employ distance metrics other than the Euclidean distances in this scheme, whic... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Proximal Classification Method based on Two Smallest and Supervised Hyperspheres

    Publication Year: 2007, Page(s):63 - 68
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (811 KB) | HTML iconHTML

    We propose a proximal classification method, named as the hyperspherical 2-surface proximal (H2SP) classifier, by seeking the two smallest hyperspheres for the positive class and the negative class, respectively, each containing the most samples from one class while also the least samples from the other. The proposed H2SP classifier is validated using five public benchmark datasets, including one ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Feature Selection for Non Gaussian Mixture Models

    Publication Year: 2007, Page(s):69 - 74
    Cited by:  Papers (2)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (1336 KB) | HTML iconHTML

    We present in this paper a new approach for unsupervised feature selection for non Gaussian data controlled by a finite mixture of generalized Dirichlet distributions. We model each feature by a mixture of two Beta distributions: one relevant and depends on component labels while the second is uninformative for the clustering. The relevance of each feature is then quantified by the mixture weight ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Scalable, Efficient, Stepwise-Optimal Feature Elimination in Support Vector Machines

    Publication Year: 2007, Page(s):75 - 80
    Cited by:  Papers (2)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (573 KB) | HTML iconHTML

    We address feature selection for support vector machines for the scenario in which the feature space is huge, i.e., 105 - 106 or more features, as may occur e.g. in a biomedical context working with 3-D (or 4-D) brain images. Feature selection in this case may be needed to improve the classifier's generalization performance (given limited training data), to reduce classificat... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Learning with Heterogenous Data Sets by Weighted Multiple Kernel Canonical Correlation Analysis

    Publication Year: 2007, Page(s):81 - 86
    Cited by:  Papers (2)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (4263 KB) | HTML iconHTML

    A new formulation of weighted multiple kernel based canonical correlation analysis(WMKCCA) is proposed in this paper. Computational issues are also considered in the proposed method to make it feasible on large data sets. This method uses incomplete Cholesky decomposition(ICD) and singular value decomposition(S VD) to approximate the original eigenvalue problem for low rank. For the weighted exten... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Multi-Objective Programming Approach to Compromising Classification Performance Metrics

    Publication Year: 2007, Page(s):87 - 92
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (1694 KB) | HTML iconHTML

    In this paper, we propose an MOP approach for finding the best compromise solution among more than two competing performance criteria. Our formulation for classifier learning, which we refer to as iterative constrained optimization (ICO), involves an iterative process of the optimization of individual objectives with proper constraints on the remaining competing objectives. The fundamental idea is... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Bayesian Image Restoration Based On Variatonal Inference and a Product of Student-t Priors

    Publication Year: 2007, Page(s):93 - 98
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (1258 KB) | HTML iconHTML

    Image priors based on products have been recognized to offer many advantages since they provide the ability to enforce simultaneously multiple constraints. However, they are inconvenient for Bayesian inference since their normalization constant cannot be found in closed form. In this paper a new Bayesian framework is proposed for the image restoration problem, where the observed image is degraded ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Pit Pattern Classification of Zoom-Endoscopical Colon Images using Evolved Fourier Feature Vectors

    Publication Year: 2007, Page(s):99 - 104
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (3257 KB) | HTML iconHTML

    This work describes an experimental study on the classification of images taken from colonoscopy. An emphasis is devoted to the procedure of finding features which allow an adequate classification. The proposed approach applies filters to the images' respective Fourier domains. Good configurations of these filters are obtained using a genetic algorithm, since the complexity of the configuration sp... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • 3-D Scene Modelling from Multiple Images using Radial Basis Function Networks

    Publication Year: 2007, Page(s):105 - 110
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (2688 KB) | HTML iconHTML

    A new approach for modelling multiple 3-D objects from a sparse set of images taken from various viewpoints is proposed in this paper. A voxel model of the scene is estimated from the given set of images using the space carving algorithm. An implicit radial basis function (RBF) network is used afterwards to model the voxel data. The multiorder function is chosen as the kernel function due to its p... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.