Efficient algorithms for learning kernels from multiple similarity matrices with general convex loss functions

In this paper we consider the problem of learning an n × n kernel matrix from m(1) similarity matrices under general convex loss. Past research have extensively studied the m = 1 case and have derived several algorithms which require sophisticated techniques like ACCP, SOCP, etc. The existing algorithms do not apply if one uses arbitrary losses and often can not handle m > 1 case. We present several provably convergent iterative algorithms, where each iteration requires either an SVM or a Multiple Kernel Learning (MKL) solver for m > 1 case. One of the major contributions of the paper is to extend the well knownMirror Descent(MD) framework to handle Cartesian product of psd matrices. This novel extension leads to an algorithm, called EMKL, which solves the problem in O(m2 log n 2) iterations; in each iteration one solves an MKL involving m kernels and m eigen-decomposition of n × n matrices. By suitably defining a restriction on the objective function, a faster version of EMKL is proposed, called REKL,which avoids the eigen-decomposition. An alternative to both EMKL and REKL is also suggested which requires only an SVMsolver. Experimental results on real world protein data set involving several similarity matrices illustrate the efficacy of the proposed algorithms.

A Huber-loss-driven clustering technique and its application to robust cell detection in confocal microscopy images

We address the problem of detecting cells in biological images. The problem is important in many automated image analysis applications. We identify the problem as one of clustering and formulate it within the framework of robust estimation using loss functions. We show how suitable loss functions may be chosen based on a priori knowledge of the noise distribution. Specifically, in the context of biological images, since the measurement noise is not Gaussian, quadratic loss functions yield suboptimal results. We show that by incorporating the Huber loss function, cells can be detected robustly and accurately. To initialize the algorithm, we also propose a seed selection approach. Simulation results show that Huber loss exhibits better performance compared with some standard loss functions. We also provide experimental results on confocal images of yeast cells. The proposed technique exhibits good detection performance even when the signal-to-noise ratio is low.

Structured Learning for Non-Smooth Ranking losses

Learning to rank from relevance judgment is an active research area. Itemwise score regression, pairwise preference satisfaction, and listwise structured learning are the major techniques in use. Listwise structured learning has been applied recently to optimize important non-decomposable ranking criteria like AUC (area under ROC curve) and MAP(mean average precision). We propose new, almost-lineartime algorithms to optimize for two other criteria widely used to evaluate search systems: MRR (mean reciprocal rank) and NDCG (normalized discounted cumulative gain)in the max-margin structured learning framework. We also demonstrate that, for different ranking criteria, one may need to use different feature maps. Search applications should not be optimized in favor of a single criterion, because they need to cater to a variety of queries. E.g., MRR is best for navigational queries, while NDCG is best for informational queries. A key contribution of this paper is to fold multiple ranking loss functions into a multi-criteria max-margin optimization.The result is a single, robust ranking model that is close to the best accuracy of learners trained on individual criteria. In fact, experiments over the popular LETOR and TREC data sets show that, contrary to conventional wisdom, a test criterion is often not best served by training with the same individual criterion.

Noise tolerance under risk minimization

In this paper, we explore noise-tolerant learning of classifiers. We formulate the problem as follows. We assume that there is an unobservable training set that is noise free. The actual training set given to the learning algorithm is obtained from this ideal data set by corrupting the class label of each example. The probability that the class label of an example is corrupted is a function of the feature vector of the example. This would account for most kinds of noisy data one encounters in practice. We say that a learning method is noise tolerant if the classifiers learnt with noise-free data and with noisy data, both have the same classification accuracy on the noise-free data. In this paper, we analyze the noise-tolerance properties of risk minimization (under different loss functions). We show that risk minimization under 0-1 loss function has impressive noise-tolerance properties and that under squared error loss is tolerant only to uniform noise; risk minimization under other loss functions is not noise tolerant. We conclude this paper with some discussion on the implications of these theoretical results.

Classification calibration dimension for general multiclass losses

We study consistency properties of surrogate loss functions for general multiclass classification problems, defined by a general loss matrix. We extend the notion of classification calibration, which has been studied for binary and multiclass 0-1 classification problems (and for certain other specific learning problems), to the general multiclass setting, and derive necessary and sufficient conditions for a surrogate loss to be classification calibrated with respect to a loss matrix in this setting. We then introduce the notion of \emph{classification calibration dimension} of a multiclass loss matrix, which measures the smallest `size' of a prediction space for which it is possible to design a convex surrogate that is classification calibrated with respect to the loss matrix. We derive both upper and lower bounds on this quantity, and use these results to analyze various loss matrices. In particular, as one application, we provide a different route from the recent result of Duchi et al.\ (2010) for analyzing the difficulty of designing `low-dimensional' convex surrogates that are consistent with respect to pairwise subset ranking losses. We anticipate the classification calibration dimension may prove to be a useful tool in the study and design of surrogate losses for general multiclass learning problems.

Temporal Envelope Fit of Transient Audio Signals

We address the problem of temporal envelope modeling for transient audio signals. We propose the Gamma distribution function (GDF) as a suitable candidate for modeling the envelope keeping in view some of its interesting properties such as asymmetry, causality, near-optimal time-bandwidth product, controllability of rise and decay, etc. The problem of finding the parameters of the GDF becomes a nonlinear regression problem. We overcome the hurdle by using a logarithmic envelope fit, which reduces the problem to one of linear regression. The logarithmic transformation also has the feature of dynamic range compression. Since temporal envelopes of audio signals are not uniformly distributed, in order to compute the amplitude, we investigate the importance of various loss functions for regression. Based on synthesized data experiments, wherein we have a ground truth, and real-world signals, we observe that the least-squares technique gives reasonably accurate amplitude estimates compared with other loss functions.

Extension of TSVM to multi-class and hierarchical text classification problems With general losses

Transductive SVM (TSVM) is a well known semi-supervised large margin learning method for binary text classification. In this paper we extend this method to multi-class and hierarchical classification problems. We point out that the determination of labels of unlabeled examples with fixed classifier weights is a linear programming problem. We devise an efficient technique for solving it. The method is applicable to general loss functions. We demonstrate the value of the new method using large margin loss on a number of multi-class and hierarchical classification datasets. For maxent loss we show empirically that our method is better than expectation regularization/constraint and posterior regularization methods, and competitive with the version of entropy regularization method which uses label constraints.

Making risk minimization tolerant to label noise

In many applications, the training data, from which one needs to learn a classifier, is corrupted with label noise. Many standard algorithms such as SVM perform poorly in the presence of label noise. In this paper we investigate the robustness of risk minimization to label noise. We prove a sufficient condition on a loss function for the risk minimization under that loss to be tolerant to uniform label noise. We show that the 0-1 loss, sigmoid loss, ramp loss and probit loss satisfy this condition though none of the standard convex loss functions satisfy it. We also prove that, by choosing a sufficiently large value of a parameter in the loss function, the sigmoid loss, ramp loss and probit loss can be made tolerant to nonuniform label noise also if we can assume the classes to be separable under noise-free data distribution. Through extensive empirical studies, we show that risk minimization under the 0-1 loss, the sigmoid loss and the ramp loss has much better robustness to label noise when compared to the SVM algorithm. (C) 2015 Elsevier B.V. All rights reserved.

Transmission loss analysis of rectangular expansion chamber with arbitrary location of inlet/outlet by means of Green's functions

Transmission loss of a rectangular expansion chamber, the inlet and outlet of which are situated at arbitrary locations of the chamber, i.e., the side wall or the face of the chamber, are analyzed here based on the Green's function of a rectangular cavity with homogeneous boundary conditions. The rectangular chamber Green's function is expressed in terms of a finite number of rigid rectangular cavity mode shapes. The inlet and outlet ports are modeled as uniform velocity pistons. If the size of the piston is small compared to wavelength, then the plane wave excitation is a valid assumption. The velocity potential inside the chamber is expressed by superimposing the velocity potentials of two different configurations. The first configuration is a piston source at the inlet port and a rigid termination at the outlet, and the second one is a piston at the outlet with a rigid termination at the inlet. Pressure inside the chamber is derived from velocity potentials using linear momentum equation. The average pressure acting on the pistons at the inlet and outlet locations is estimated by integrating the acoustic pressure over the piston area in the two constituent configurations. The transfer matrix is derived from the average pressure values and thence the transmission loss is calculated. The results are verified against those in the literature where use has been made of modal expansions and also numerical models (FEM fluid). The transfer matrix formulation for yielding wall rectangular chambers has been derived incorporating the structural–acoustic coupling. Parametric studies are conducted for different inlet and outlet configurations, and the various phenomena occurring in the TL curves that cannot be explained by the classical plane wave theory, are discussed.

Glycodelin-A interferes with IL-2/IL-2R signalling to induce cell growth arrest, loss of effector functions and apoptosis in T-lymphocytes

The progesterone-regulated glycoprotein glycodelin-A (GdA), secreted by the decidualized endometrium at high concentrations in primates, inhibits the maternal immune response against fetal antigens and thereby contributes to the tolerance of the semi-allogenic fetus during a normal pregnancy. Our earlier studies demonstrated the ability of GdA to induce an intrinsic apoptotic cascade in CD4 T-lymphocytes and suppress the cytolytic effector function of CD8 T-lymphocytes. In this report, we investigated further into the mechanism of action of GdA controlling perforin and granzyme B expression in CD8 T-lymphocytes and the mechanism of action of GdA leading to lymphocyte death. Flow cytometry analysis was performed to check for the surface expression of interleukin-2 receptor (IL-2R) and intracellular eomesodermin (Eomes) in activated T-lymphocytes, whereas quantitative RTPCR analysis was used to find out their mRNA profile upon GdA treatment. Western analysis was carried out to confirm the protein level of Bax and Bcl-2. GdA reduces the surface expression of the high-affinity IL-2R complex by down-regulating the synthesis of IL-2R (CD25). This disturbs the optimal IL-2 signalling and decreases the Eomes expression, which along with IL-2 directly regulates perforin and granzymes expression. Consequently, the CD8 T-lymphocytes undergo growth arrest and are unable to mature into competent cytotoxic T-lymphocytes. In the CD4 T-lymphocytes, growth factor IL-2 deprivation leads to proliferation inhibition, decreased Bcl-2/enhanced Bax expression, culminating in mitochondrial stress and cell death. GdA spurs cell cycle arrest, loss of effector functions and apoptosis in different T-cell subsets by making T-lymphocytes unable to respond to IL-2.

Functions Of Cytochrome-C In Regulation Of Electron-Transfer And Protein Folding

Cytochrome c, a "mobile electron carrier" of the mitochondrial respiratory chain, also occurs in detectable amounts in the cytosol, and can receive electrons from cytochromes present in endoplasmic reticulum and plasma membranes as well as from superoxide and ascorbate. The pigment was found to dissociate from mitochondrial membranes in liver and kidney when rats were subjected to heat exposure and starvation, respectively. Treating cytochrome c with hydroxylamine gives a partially deaminated product with altered redox properties; decreased stimulation of respiration by deficient mitochondria, increased reduction by superoxide, and complete loss of reducibility by plasma membranes. Mitochondria isolated from brown adipose tissue of cold-exposed rats are found to be sub-saturated with cytochrome c. The ability of cytochrome c to reactivate reduced ribonuclease is now reinterpreted as a molecular chaperone role for the hemoprotein.

Dielectric function and optical conductivity of TiOx (0.8loss spectroscopy

Reflection electron energy-loss spectra are reported for the family of compounds TiOx over the entire homogeneity range (0.8 < a: < 1.3). The spectra exhibit a plasmon feature on the low-energy side, while several interband transitions are prominent at higher energies. The real and imaginary parts of dielectric functions and optical conductivity for these compounds are determined using the Kramers-Kronig analysis. The results exhibit systematic behavior with varying oxygen stoichiometry.

Typical and atypical domain combinations in human protein kinases: functions, disease causing mutations and conservation in other primates

Ser/Thr and Tyr protein kinases orchestrate many signalling pathways and hence loss in this balance leads to many disease phenotypes. Due to their high abundance, diversity and importance, efforts have been made in the past to classify kinases and annotate their functions at both gross and fine levels. These kinases are conventionally classified into subfamilies based on the sequences of catalytic domains. Usually the domain architecture of a full-length kinase is consistent with the subfamily classification made based on the sequence of kinase domain. Important contributions of modular domains to the overall function of the kinase are well known. Recently occurrence of two kinds of outlier kinases-''Hybrid'' and ``Rogue'' has been reported. These show considerable deviations in their domain architectures from the typical domain architecture known for the classical kinase subfamilies. This article provides an overview of the different subfamilies of human kinases and the role of non-kinase domains in functions and diseases. Importantly this article provides analysis of hybrid and rogue kinases encoded in the human genome and highlights their conservation in closely related primate species. These kinases are examples of elegant rewiring to bring about subtle functional differences compared to canonical variants.

Finite volume evolution Galerkin method for hyperbolic conservation laws with spatially varying flux functions

We present a generalization of the finite volume evolution Galerkin scheme [M. Lukacova-Medvid'ova,J. Saibertov'a, G. Warnecke, Finite volume evolution Galerkin methods for nonlinear hyperbolic systems, J. Comp. Phys. (2002) 183 533-562; M. Luacova-Medvid'ova, K.W. Morton, G. Warnecke, Finite volume evolution Galerkin (FVEG) methods for hyperbolic problems, SIAM J. Sci. Comput. (2004) 26 1-30] for hyperbolic systems with spatially varying flux functions. Our goal is to develop a genuinely multi-dimensional numerical scheme for wave propagation problems in a heterogeneous media. We illustrate our methodology for acoustic waves in a heterogeneous medium but the results can be generalized to more complex systems. The finite volume evolution Galerkin (FVEG) method is a predictor-corrector method combining the finite volume corrector step with the evolutionary predictor step. In order to evolve fluxes along the cell interfaces we use multi-dimensional approximate evolution operator. The latter is constructed using the theory of bicharacteristics under the assumption of spatially dependent wave speeds. To approximate heterogeneous medium a staggered grid approach is used. Several numerical experiments for wave propagation with continuous as well as discontinuous wave speeds confirm the robustness and reliability of the new FVEG scheme.

Dynamics of a dilute sheared inelastic fluid. I. Hydrodynamic modes and velocity correlation functions

The hydrodynamic modes and the velocity autocorrelation functions for a dilute sheared inelastic fluid are analyzed using an expansion in the parameter epsilon=(1-e)(1/2), where e is the coefficient of restitution. It is shown that the hydrodynamic modes for a sheared inelastic fluid are very different from those for an elastic fluid in the long-wave limit, since energy is not a conserved variable when the wavelength of perturbations is larger than the ``conduction length.'' In an inelastic fluid under shear, there are three coupled modes, the mass and the momenta in the plane of shear, which have a decay rate proportional to k(2/3) in the limit k -> 0, if the wave vector has a component along the flow direction. When the wave vector is aligned along the gradient-vorticity plane, we find that the scaling of the growth rate is similar to that for an elastic fluid. The Fourier transforms of the velocity autocorrelation functions are calculated for a steady shear flow correct to leading order in an expansion in epsilon. The time dependence of the autocorrelation function in the long-time limit is obtained by estimating the integral of the Fourier transform over wave number space. It is found that the autocorrelation functions for the velocity in the flow and gradient directions decay proportional to t(-5/2) in two dimensions and t(-15/4) in three dimensions. In the vorticity direction, the decay of the autocorrelation function is proportional to t(-3) in two dimensions and t(-7/2) in three dimensions.