Header logo is ei


2002


no image
A compression approach to support vector model selection

von Luxburg, U., Bousquet, O., Schölkopf, B.

(101), Max Planck Institute for Biological Cybernetics, 2002, see more detailed JMLR version (techreport)

Abstract
In this paper we investigate connections between statistical learning theory and data compression on the basis of support vector machine (SVM) model selection. Inspired by several generalization bounds we construct ``compression coefficients'' for SVMs, which measure the amount by which the training labels can be compressed by some classification hypothesis. The main idea is to relate the coding precision of this hypothesis to the width of the margin of the SVM. The compression coefficients connect well known quantities such as the radius-margin ratio R^2/rho^2, the eigenvalues of the kernel matrix and the number of support vectors. To test whether they are useful in practice we ran model selection experiments on several real world datasets. As a result we found that compression coefficients can fairly accurately predict the parameters for which the test error is minimized.

[BibTex]

2002

[BibTex]


no image
A Bennett Concentration Inequality and Its Application to Suprema of Empirical Processes

Bousquet, O.

C. R. Acad. Sci. Paris, Ser. I, 334, pages: 495-500, 2002 (article)

Abstract
We introduce new concentration inequalities for functions on product spaces. They allow to obtain a Bennett type deviation bound for suprema of empirical processes indexed by upper bounded functions. The result is an improvement on Rio's version \cite{Rio01b} of Talagrand's inequality \cite{Talagrand96} for equidistributed variables.

PDF PostScript [BibTex]


no image
Numerical evolution of axisymmetric, isolated systems in general relativity

Frauendiener, J., Hein, M.

Physical Review D, 66, pages: 124004-124004, 2002 (article)

Abstract
We describe in this article a new code for evolving axisymmetric isolated systems in general relativity. Such systems are described by asymptotically flat space-times, which have the property that they admit a conformal extension. We are working directly in the extended conformal manifold and solve numerically Friedrich's conformal field equations, which state that Einstein's equations hold in the physical space-time. Because of the compactness of the conformal space-time the entire space-time can be calculated on a finite numerical grid. We describe in detail the numerical scheme, especially the treatment of the axisymmetry and the boundary.

GZIP [BibTex]

GZIP [BibTex]


no image
Marginalized kernels for biological sequences

Tsuda, K., Kin, T., Asai, K.

Bioinformatics, 18(Suppl 1):268-275, 2002 (article)

PDF [BibTex]

PDF [BibTex]


no image
Localized Rademacher Complexities

Bartlett, P., Bousquet, O., Mendelson, S.

In Proceedings of the 15th annual conference on Computational Learning Theory, pages: 44-58, Proceedings of the 15th annual conference on Computational Learning Theory, 2002 (inproceedings)

Abstract
We investigate the behaviour of global and local Rademacher averages. We present new error bounds which are based on the local averages and indicate how data-dependent local averages can be estimated without {it a priori} knowledge of the class at hand.

PDF PostScript [BibTex]

PDF PostScript [BibTex]


no image
Film Cooling: A Comparative Study of Different Heaterfoil Configurations for Liquid Crystals Experiments

Vogel, G., Graf, ABA., Weigand, B.

In ASME TURBO EXPO 2002, Amsterdam, GT-2002-30552, ASME TURBO EXPO, Amsterdam, 2002 (inproceedings)

PDF [BibTex]

PDF [BibTex]


no image
Feature Selection and Transduction for Prediction of Molecular Bioactivity for Drug Design

Weston, J., Perez-Cruz, F., Bousquet, O., Chapelle, O., Elisseeff, A., Schölkopf, B.

Max Planck Institute for Biological Cybernetics / Biowulf Technologies, 2002 (techreport)

Web [BibTex]

Web [BibTex]


no image
Application of Monte Carlo Methods to Psychometric Function Fitting

Wichmann, F.

Proceedings of the 33rd European Conference on Mathematical Psychology, pages: 44, 2002 (poster)

Abstract
The psychometric function relates an observer's performance to an independent variable, usually some physical quantity of a stimulus in a psychophysical task. Here I describe methods to (1) fitting psychometric functions, (2) assessing goodness-of-fit, and (3) providing confidence intervals for the function's parameters and other estimates derived from them. First I describe a constrained maximum-likelihood method for parameter estimation. Using Monte-Carlo simulations I demonstrate that it is important to have a fitting method that takes stimulus-independent errors (or "lapses") into account. Second, a number of goodness-of-fit tests are introduced. Because psychophysical data sets are usually rather small I advocate the use of Monte Carlo resampling techniques that do not rely on asymptotic theory for goodness-of-fit assessment. Third, a parametric bootstrap is employed to estimate the variability of fitted parameters and derived quantities such as thresholds and slopes. I describe how the bootstrap bridging assumption, on which the validity of the procedure depends, can be tested without incurring too high a cost in computation time. Finally I describe how the methods can be extended to test hypotheses concerning the form and shape of several psychometric functions. Software describing the methods is available (http://www.bootstrap-software.com/psignifit/), as well as articles describing the methods in detail (Wichmann&Hill, Perception&Psychophysics, 2001a,b).

[BibTex]

[BibTex]


no image
Stability and Generalization

Bousquet, O., Elisseeff, A.

Journal of Machine Learning Research, 2, pages: 499-526, 2002 (article)

Abstract
We define notions of stability for learning algorithms and show how to use these notions to derive generalization error bounds based on the empirical error and the leave-one-out error. The methods we use can be applied in the regression framework as well as in the classification one when the classifier is obtained by thresholding a real-valued function. We study the stability properties of large classes of learning algorithms such as regularization based algorithms. In particular we focus on Hilbert space regularization and Kullback-Leibler regularization. We demonstrate how to apply the results to SVM for regression and classification.

PDF PostScript [BibTex]

PDF PostScript [BibTex]


no image
Subspace information criterion for non-quadratic regularizers – model selection for sparse regressors

Tsuda, K., Sugiyama, M., Müller, K.

IEEE Trans Neural Networks, 13(1):70-80, 2002 (article)

PDF [BibTex]

PDF [BibTex]


no image
Modeling splicing sites with pairwise correlations

Arita, M., Tsuda, K., Asai, K.

Bioinformatics, 18(Suppl 2):27-34, 2002 (article)

PDF [BibTex]

PDF [BibTex]


no image
Observations on the Nyström Method for Gaussian Process Prediction

Williams, C., Rasmussen, C., Schwaighofer, A., Tresp, V.

Max Planck Institute for Biological Cybernetics, Tübingen, Germany, 2002 (techreport)

Abstract
A number of methods for speeding up Gaussian Process (GP) prediction have been proposed, including the Nystr{\"o}m method of Williams and Seeger (2001). In this paper we focus on two issues (1) the relationship of the Nystr{\"o}m method to the Subset of Regressors method (Poggio and Girosi 1990; Luo and Wahba, 1997) and (2) understanding in what circumstances the Nystr{\"o}m approximation would be expected to provide a good approximation to exact GP regression.

PostScript [BibTex]

PostScript [BibTex]


no image
Perfusion Quantification using Gaussian Process Deconvolution

Andersen, IK., Szymkowiak, A., Rasmussen, CE., Hanson, LG., Marstrand, JR., Larsson, HBW., Hansen, LK.

Magnetic Resonance in Medicine, (48):351-361, 2002 (article)

Abstract
The quantification of perfusion using dynamic susceptibility contrast MR imaging requires deconvolution to obtain the residual impulse-response function (IRF). Here, a method using a Gaussian process for deconvolution, GPD, is proposed. The fact that the IRF is smooth is incorporated as a constraint in the method. The GPD method, which automatically estimates the noise level in each voxel, has the advantage that model parameters are optimized automatically. The GPD is compared to singular value decomposition (SVD) using a common threshold for the singular values and to SVD using a threshold optimized according to the noise level in each voxel. The comparison is carried out using artificial data as well as using data from healthy volunteers. It is shown that GPD is comparable to SVD variable optimized threshold when determining the maximum of the IRF, which is directly related to the perfusion. GPD provides a better estimate of the entire IRF. As the signal to noise ratio increases or the time resolution of the measurements increases, GPD is shown to be superior to SVD. This is also found for large distribution volumes.

PDF PostScript [BibTex]

PDF PostScript [BibTex]


no image
Tracking a Small Set of Experts by Mixing Past Posteriors

Bousquet, O., Warmuth, M.

Journal of Machine Learning Research, 3, pages: 363-396, (Editors: Long, P.), 2002 (article)

Abstract
In this paper, we examine on-line learning problems in which the target concept is allowed to change over time. In each trial a master algorithm receives predictions from a large set of n experts. Its goal is to predict almost as well as the best sequence of such experts chosen off-line by partitioning the training sequence into k+1 sections and then choosing the best expert for each section. We build on methods developed by Herbster and Warmuth and consider an open problem posed by Freund where the experts in the best partition are from a small pool of size m. Since k >> m, the best expert shifts back and forth between the experts of the small pool. We propose algorithms that solve this open problem by mixing the past posteriors maintained by the master algorithm. We relate the number of bits needed for encoding the best partition to the loss bounds of the algorithms. Instead of paying log n for choosing the best expert in each section we first pay log (n choose m) bits in the bounds for identifying the pool of m experts and then log m bits per new section. In the bounds we also pay twice for encoding the boundaries of the sections.

PDF PostScript [BibTex]

PDF PostScript [BibTex]


no image
A femoral arteriovenous shunt facilitates arterial whole blood sampling in animals

Weber, B., Burger, C., Biro, P., Buck, A.

Eur J Nucl Med Mol Imaging, 29, pages: 319-323, 2002 (article)

[BibTex]

[BibTex]


no image
Some Local Measures of Complexity of Convex Hulls and Generalization Bounds

Bousquet, O., Koltchinskii, V., Panchenko, D.

In Proceedings of the 15th annual conference on Computational Learning Theory, Proceedings of the 15th annual conference on Computational Learning Theory, 2002 (inproceedings)

Abstract
We investigate measures of complexity of function classes based on continuity moduli of Gaussian and Rademacher processes. For Gaussian processes, we obtain bounds on the continuity modulus on the convex hull of a function class in terms of the same quantity for the class itself. We also obtain new bounds on generalization error in terms of localized Rademacher complexities. This allows us to prove new results about generalization performance for convex hulls in terms of characteristics of the base class. As a byproduct, we obtain a simple proof of some of the known bounds on the entropy of convex hulls.

PDF PostScript [BibTex]

PDF PostScript [BibTex]


no image
Contrast discrimination with pulse-trains in pink noise

Henning, G., Bird, C., Wichmann, F.

Journal of the Optical Society of America A, 19(7), pages: 1259-1266, 2002 (article)

Abstract
Detection performance was measured with sinusoidal and pulse-train gratings. Although the 2.09-c/deg pulse-train, or line gratings, contained at least 8 harmonics all at equal contrast, they were no more detectable than their most detectable component. The addition of broadband pink noise designed to equalize the detectability of the components of the pulse train made the pulse train about a factor of four more detectable than any of its components. However, in contrast-discrimination experiments, with a pedestal or masking grating of the same form and phase as the signal and 15% contrast, the noise did not affect the discrimination performance of the pulse train relative to that obtained with its sinusoidal components. We discuss the implications of these observations for models of early vision in particular the implications for possible sources of internal noise.

PDF [BibTex]

PDF [BibTex]


no image
A kernel approach for learning from almost orthogonal patterns

Schölkopf, B., Weston, J., Eskin, E., Leslie, C., Noble, W.

In Principles of Data Mining and Knowledge Discovery, Lecture Notes in Computer Science, 2430/2431, pages: 511-528, Lecture Notes in Computer Science, (Editors: T Elomaa and H Mannila and H Toivonen), Springer, Berlin, Germany, 13th European Conference on Machine Learning (ECML) and 6th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD'2002), 2002 (inproceedings)

PostScript DOI [BibTex]

PostScript DOI [BibTex]


no image
Optimal linear estimation of self-motion - a real-world test of a model of fly tangential neurons

Franz, MO.

SAB 02 Workshop, Robotics as theoretical biology, 7th meeting of the International Society for Simulation of Adaptive Behaviour (SAB), (Editors: Prescott, T.; Webb, B.), 2002 (poster)

Abstract
The tangential neurons in the fly brain are sensitive to the typical optic flow patterns generated during self-motion (see example in Fig.1). We examine whether a simplified linear model of these neurons can be used to estimate self-motion from the optic flow. We present a theory for the construction of an optimal linear estimator incorporating prior knowledge both about the distance distribution of the environment, and about the noise and self-motion statistics of the sensor. The optimal estimator is tested on a gantry carrying an omnidirectional vision sensor that can be moved along three translational and one rotational degree of freedom. The experiments indicate that the proposed approach yields accurate results for rotation estimates, independently of the current translation and scene layout. Translation estimates, however, turned out to be sensitive to simultaneous rotation and to the particular distance distribution of the scene. The gantry experiments confirm that the receptive field organization of the tangential neurons allows them, as an ensemble, to extract self-motion from the optic flow.

PDF [BibTex]

PDF [BibTex]


no image
Choosing Multiple Parameters for Support Vector Machines

Chapelle, O., Vapnik, V., Bousquet, O., Mukherjee, S.

Machine Learning, 46(1):131-159, 2002 (article)

Abstract
The problem of automatically tuning multiple parameters for pattern recognition Support Vector Machines (SVM) is considered. This is done by minimizing some estimates of the generalization error of SVMs using a gradient descent algorithm over the set of parameters. Usual methods for choosing parameters, based on exhaustive search become intractable as soon as the number of parameters exceeds two. Some experimental results assess the feasibility of our approach for a large number of parameters (more than 100) and demonstrate an improvement of generalization performance.

PDF PostScript [BibTex]

PDF PostScript [BibTex]


no image
Infinite Mixtures of Gaussian Process Experts

Rasmussen, CE., Ghahramani, Z.

In (Editors: Dietterich, Thomas G.; Becker, Suzanna; Ghahramani, Zoubin), 2002 (inproceedings)

Abstract
We present an extension to the Mixture of Experts (ME) model, where the individual experts are Gaussian Process (GP) regression models. Using a input-dependent adaptation of the Dirichlet Process, we implement a gating network for an infinite number of Experts. Inference in this model may be done efficiently using a Markov Chain relying on Gibbs sampling. The model allows the effective covariance function to vary with the inputs, and may handle large datasets -- thus potentially overcoming two of the biggest hurdles with GP models. Simulations show the viability of this approach.

PDF PostScript [BibTex]

PDF PostScript [BibTex]


no image
Marginalized kernels for RNA sequence data analysis

Kin, T., Tsuda, K., Asai, K.

In Genome Informatics 2002, pages: 112-122, (Editors: Lathtop, R. H.; Nakai, K.; Miyano, S.; Takagi, T.; Kanehisa, M.), Genome Informatics, 2002, (Best Paper Award) (inproceedings)

Web [BibTex]

Web [BibTex]


no image
Luminance Artifacts on CRT Displays

Wichmann, F.

In IEEE Visualization, pages: 571-574, (Editors: Moorhead, R.; Gross, M.; Joy, K. I.), IEEE Visualization, 2002 (inproceedings)

Abstract
Most visualization panels today are still built around cathode-ray tubes (CRTs), certainly on personal desktops at work and at home. Whilst capable of producing pleasing images for common applications ranging from email writing to TV and DVD presentation, it is as well to note that there are a number of nonlinear transformations between input (voltage) and output (luminance) which distort the digital and/or analogue images send to a CRT. Some of them are input-independent and hence easy to fix, e.g. gamma correction, but others, such as pixel interactions, depend on the content of the input stimulus and are thus harder to compensate for. CRT-induced image distortions cause problems not only in basic vision research but also for applications where image fidelity is critical, most notably in medicine (digitization of X-ray images for diagnostic purposes) and in forms of online commerce, such as the online sale of images, where the image must be reproduced on some output device which will not have the same transfer function as the customer's CRT. I will present measurements from a number of CRTs and illustrate how some of their shortcomings may be problematic for the aforementioned applications.

[BibTex]

[BibTex]

2001


no image
Pattern Selection Using the Bias and Variance of Ensemble

Shin, H., Cho, S.

In Proc. of the Korean Data Mining Conference, pages: 56-67, Korean Data Mining Conference, December 2001 (inproceedings)

[BibTex]

2001

[BibTex]


no image
Separation of post-nonlinear mixtures using ACE and temporal decorrelation

Ziehe, A., Kawanabe, M., Harmeling, S., Müller, K.

In ICA 2001, pages: 433-438, (Editors: Lee, T.-W. , T.P. Jung, S. Makeig, T. J. Sejnowski), Third International Workshop on Independent Component Analysis and Blind Signal Separation, December 2001 (inproceedings)

Abstract
We propose an efficient method based on the concept of maximal correlation that reduces the post-nonlinear blind source separation problem (PNL BSS) to a linear BSS problem. For this we apply the Alternating Conditional Expectation (ACE) algorithm – a powerful technique from nonparametric statistics – to approximately invert the (post-)nonlinear functions. Interestingly, in the framework of the ACE method convergence can be proven and in the PNL BSS scenario the optimal transformation found by ACE will coincide with the desired inverse functions. After the nonlinearities have been removed by ACE, temporal decorrelation (TD) allows us to recover the source signals. An excellent performance underlines the validity of our approach and demonstrates the ACE-TD method on realistic examples.

PDF [BibTex]

PDF [BibTex]


no image
Perception of Planar Shapes in Depth

Wichmann, F., Willems, B., Rosas, P., Wagemans, J.

Journal of Vision, 1(3):176, First Annual Meeting of the Vision Sciences Society (VSS), December 2001 (poster)

Abstract
We investigated the influence of the perceived 3D-orientation of planar elliptical shapes on the perception of the shapes themselves. Ellipses were projected onto the surface of a sphere and subjects were asked to indicate if the projected shapes looked as if they were a circle on the surface of the sphere. The image of the sphere was obtained from a real, (near) perfect sphere using a highly accurate digital camera (real sphere diameter 40 cm; camera-to-sphere distance 320 cm; for details see Willems et al., Perception 29, S96, 2000; Photometrics SenSys 400 digital camera with Rodenstock lens, 12-bit linear luminance resolution). Stimuli were presented monocularly on a carefully linearized Sony GDM-F500 monitor keeping the scene geometry as in the real case (sphere diameter on screen 8.2 cm; viewing distance 66 cm). Experiments were run in a darkened room using a viewing tube to minimize, as far as possible, extraneous monocular cues to depth. Three different methods were used to obtain subjects' estimates of 3D-shape: the method of adjustment, temporal 2-alternative forced choice (2AFC) and yes/no. Several results are noteworthy. First, mismatch between perceived and objective slant tended to decrease with increasing objective slant. Second, the variability of the settings, too, decreased with increasing objective slant. Finally, we comment on the results obtained using different psychophysical methods and compare our results to those obtained using a real sphere and binocular vision (Willems et al.).

Web DOI [BibTex]

Web DOI [BibTex]


no image
Anabolic and Catabolic Gene Expression Pattern Analysis in Normal Versus Osteoarthritic Cartilage Using Complementary DNA-Array Technology

Aigner, T., Zien, A., Gehrsitz, A., Gebhard, P., McKenna, L.

Arthritis and Rheumatism, 44(12):2777-2789, December 2001 (article)

Web [BibTex]

Web [BibTex]


no image
Nonlinear blind source separation using kernel feature spaces

Harmeling, S., Ziehe, A., Kawanabe, M., Blankertz, B., Müller, K.

In ICA 2001, pages: 102-107, (Editors: Lee, T.-W. , T.P. Jung, S. Makeig, T. J. Sejnowski), Third International Workshop on Independent Component Analysis and Blind Signal Separation, December 2001 (inproceedings)

Abstract
In this work we propose a kernel-based blind source separation (BSS) algorithm that can perform nonlinear BSS for general invertible nonlinearities. For our kTDSEP algorithm we have to go through four steps: (i) adapting to the intrinsic dimension of the data mapped to feature space F, (ii) finding an orthonormal basis of this submanifold, (iii) mapping the data into the subspace of F spanned by this orthonormal basis, and (iv) applying temporal decorrelation BSS (TDSEP) to the mapped data. After demixing we get a number of irrelevant components and the original sources. To find out which ones are the components of interest, we propose a criterion that allows to identify the original sources. The excellent performance of kTDSEP is demonstrated in experiments on nonlinearly mixed speech data.

PDF [BibTex]

PDF [BibTex]


no image
Pattern Selection for ‘Regression’ using the Bias and Variance of Ensemble Network

Shin, H., Cho, S.

In Proc. of the Korean Institute of Industrial Engineers Conference, pages: 10-19, Korean Industrial Engineers Conference, November 2001 (inproceedings)

[BibTex]

[BibTex]


no image
Kernel Methods for Extracting Local Image Semantics

Bradshaw, B., Schölkopf, B., Platt, J.

(MSR-TR-2001-99), Microsoft Research, October 2001 (techreport)

Web [BibTex]

Web [BibTex]


no image
Pattern Selection for ‘Classification’ using the Bias and Variance of Ensemble Neural Network

Shin, H., Cho, S.

In Proc. of the Korea Information Science Conference, pages: 307-309, Korea Information Science Conference, October 2001, Best Paper Award (inproceedings)

[BibTex]

[BibTex]


no image
Generalization performance of regularization networks and support vector machines via entropy numbers of compact operators

Williamson, R., Smola, A., Schölkopf, B.

IEEE Transactions on Information Theory, 47(6):2516-2532, September 2001 (article)

Abstract
We derive new bounds for the generalization error of kernel machines, such as support vector machines and related regularization networks by obtaining new bounds on their covering numbers. The proofs make use of a viewpoint that is apparently novel in the field of statistical learning theory. The hypothesis class is described in terms of a linear operator mapping from a possibly infinite-dimensional unit ball in feature space into a finite-dimensional space. The covering numbers of the class are then determined via the entropy numbers of the operator. These numbers, which characterize the degree of compactness of the operator can be bounded in terms of the eigenvalues of an integral operator induced by the kernel function used by the machine. As a consequence, we are able to theoretically explain the effect of the choice of kernel function on the generalization performance of support vector machines.

DOI [BibTex]

DOI [BibTex]


no image
Hybrid IDM/Impedance learning in human movements

Burdet, E., Teng, K., Chew, C., Peters, J., , B.

In ISHF 2001, 1, pages: 1-9, 1st International Symposium on Measurement, Analysis and Modeling of Human Functions (ISHF2001), September 2001 (inproceedings)

Abstract
In spite of motor output variability and the delay in the sensori-motor, humans routinely perform intrinsically un- stable tasks. The hybrid IDM/impedance learning con- troller presented in this paper enables skilful performance in strong stable and unstable environments. It consid- ers motor output variability identified from experimen- tal data, and contains two modules concurrently learning the endpoint force and impedance adapted to the envi- ronment. The simulations suggest how humans learn to skillfully perform intrinsically unstable tasks. Testable predictions are proposed.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Calibration of Digital Amateur Cameras

Urbanek, M., Horaud, R., Sturm, P.

(RR-4214), INRIA Rhone Alpes, Montbonnot, France, July 2001 (techreport)

Web [BibTex]

Web [BibTex]


no image
Combining Off- and On-line Calibration of a Digital Camera

Urbanek, M., Horaud, R., Sturm, P.

In In Proceedings of Third International Conference on 3-D Digital Imaging and Modeling, pages: 99-106, In Proceedings of Third International Conference on 3-D Digital Imaging and Modeling, June 2001 (inproceedings)

Abstract
We introduce a novel outlook on the self­calibration task, by considering images taken by a camera in motion, allowing for zooming and focusing. Apart from the complex relationship between the lens control settings and the intrinsic camera parameters, a prior off­line calibration allows to neglect the setting of focus, and to fix the principal point and aspect ratio throughout distinct views. Thus, the calibration matrix is dependent only on the zoom position. Given a fully calibrated reference view, one has only one parameter to estimate for any other view of the same scene, in order to calibrate it and to be able to perform metric reconstructions. We provide a close­form solution, and validate the reliability of the algorithm with experiments on real images. An important advantage of our method is a reduced ­ to one ­ number of critical camera configurations, associated with it. Moreover, we propose a method for computing the epipolar geometry of two views, taken from different positions and with different (spatial) resolutions; the idea is to take an appropriate third view, that is "easy" to match with the other two.

ZIP [BibTex]

ZIP [BibTex]


no image
Centralization: A new method for the normalization of gene expression data

Zien, A., Aigner, T., Zimmer, R., Lengauer, T.

Bioinformatics, 17, pages: S323-S331, June 2001, Mathematical supplement available at http://citeseer.ist.psu.edu/574280.html (article)

Abstract
Microarrays measure values that are approximately proportional to the numbers of copies of different mRNA molecules in samples. Due to technical difficulties, the constant of proportionality between the measured intensities and the numbers of mRNA copies per cell is unknown and may vary for different arrays. Usually, the data are normalized (i.e., array-wise multiplied by appropriate factors) in order to compensate for this effect and to enable informative comparisons between different experiments. Centralization is a new two-step method for the computation of such normalization factors that is both biologically better motivated and more robust than standard approaches. First, for each pair of arrays the quotient of the constants of proportionality is estimated. Second, from the resulting matrix of pairwise quotients an optimally consistent scaling of the samples is computed.

PDF PostScript Web [BibTex]

PDF PostScript Web [BibTex]


no image
Regularized principal manifolds

Smola, A., Mika, S., Schölkopf, B., Williamson, R.

Journal of Machine Learning Research, 1, pages: 179-209, June 2001 (article)

Abstract
Many settings of unsupervised learning can be viewed as quantization problems - the minimization of the expected quantization error subject to some restrictions. This allows the use of tools such as regularization from the theory of (supervised) risk minimization for unsupervised learning. This setting turns out to be closely related to principal curves, the generative topographic map, and robust coding. We explore this connection in two ways: (1) we propose an algorithm for finding principal manifolds that can be regularized in a variety of ways; and (2) we derive uniform convergence bounds and hence bounds on the learning rates of the algorithm. In particular, we give bounds on the covering numbers which allows us to obtain nearly optimal learning rates for certain types of regularization operators. Experimental results demonstrate the feasibility of the approach.

PDF [BibTex]

PDF [BibTex]


no image
Variationsverfahren zur Untersuchung von Grundzustandseigenschaften des Ein-Band Hubbard-Modells

Eichhorn, J.

Biologische Kybernetik, Technische Universität Dresden, Dresden/Germany, May 2001 (diplomathesis)

Abstract
Using different modifications of a new variational approach, statical groundstate properties of the one-band Hubbard model such as energy and staggered magnetisation are calculated. By taking into account additional fluctuations, the method ist gradually improved so that a very good description of the energy in one and two dimensions can be achieved. After a detailed discussion of the application in one dimension, extensions for two dimensions are introduced. By use of a modified version of the variational ansatz in particular a description of the quantum phase transition for the magnetisation should be possible.

PostScript [BibTex]

PostScript [BibTex]


no image
Failure Diagnosis of Discrete Event Systems

Son, HI., Kim, KW., Lee, S.

Journal of Control, Automation and Systems Engineering, 7(5):375-383, May 2001, In Korean (article)

[BibTex]

[BibTex]


no image
Support vector novelty detection applied to jet engine vibration spectra

Hayton, P., Schölkopf, B., Tarassenko, L., Anuzis, P.

In Advances in Neural Information Processing Systems 13, pages: 946-952, (Editors: TK Leen and TG Dietterich and V Tresp), MIT Press, Cambridge, MA, USA, 14th Annual Neural Information Processing Systems Conference (NIPS), April 2001 (inproceedings)

Abstract
A system has been developed to extract diagnostic information from jet engine carcass vibration data. Support Vector Machines applied to novelty detection provide a measure of how unusual the shape of a vibration signature is, by learning a representation of normality. We describe a novel method for Support Vector Machines of including information from a second class for novelty detection and give results from the application to Jet Engine vibration analysis.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Four-legged Walking Gait Control Using a Neuromorphic Chip Interfaced to a Support Vector Learning Algorithm

Still, S., Schölkopf, B., Hepp, K., Douglas, R.

In Advances in Neural Information Processing Systems 13, pages: 741-747, (Editors: TK Leen and TG Dietterich and V Tresp), MIT Press, Cambridge, MA, USA, 14th Annual Neural Information Processing Systems Conference (NIPS), April 2001 (inproceedings)

Abstract
To control the walking gaits of a four-legged robot we present a novel neuromorphic VLSI chip that coordinates the relative phasing of the robot's legs similar to how spinal Central Pattern Generators are believed to control vertebrate locomotion [3]. The chip controls the leg movements by driving motors with time varying voltages which are the outputs of a small network of coupled oscillators. The characteristics of the chip's output voltages depend on a set of input parameters. The relationship between input parameters and output voltages can be computed analytically for an idealized system. In practice, however, this ideal relationship is only approximately true due to transistor mismatch and offsets.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Algorithmic Stability and Generalization Performance

Bousquet, O., Elisseeff, A.

In Advances in Neural Information Processing Systems 13, pages: 196-202, (Editors: Leen, T.K. , T.G. Dietterich, V. Tresp), MIT Press, Cambridge, MA, USA, Fourteenth Annual Neural Information Processing Systems Conference (NIPS), April 2001 (inproceedings)

Abstract
We present a novel way of obtaining PAC-style bounds on the generalization error of learning algorithms, explicitly using their stability properties. A {\em stable} learner being one for which the learned solution does not change much for small changes in the training set. The bounds we obtain do not depend on any measure of the complexity of the hypothesis space (e.g. VC dimension) but rather depend on how the learning algorithm searches this space, and can thus be applied even when the VC dimension in infinite. We demonstrate that regularization networks possess the required stability property and apply our method to obtain new bounds on their generalization performance.

PDF Web [BibTex]

PDF Web [BibTex]


no image
The Kernel Trick for Distances

Schölkopf, B.

In Advances in Neural Information Processing Systems 13, pages: 301-307, (Editors: TK Leen and TG Dietterich and V Tresp), MIT Press, Cambridge, MA, USA, 14th Annual Neural Information Processing Systems Conference (NIPS), April 2001 (inproceedings)

Abstract
A method is described which, like the kernel trick in support vector machines (SVMs), lets us generalize distance-based algorithms to operate in feature spaces, usually nonlinearly related to the input space. This is done by identifying a class of kernels which can be represented as norm-based distances in Hilbert spaces. It turns out that the common kernel algorithms, such as SVMs and kernel PCA, are actually really distance based algorithms and can be run with that class of kernels, too. As well as providing a useful new insight into how these algorithms work, the present work can form the basis for conceiving new algorithms.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Vicinal Risk Minimization

Chapelle, O., Weston, J., Bottou, L., Vapnik, V.

In Advances in Neural Information Processing Systems 13, pages: 416-422, (Editors: Leen, T.K. , T.G. Dietterich, V. Tresp), MIT Press, Cambridge, MA, USA, Fourteenth Annual Neural Information Processing Systems Conference (NIPS) , April 2001 (inproceedings)

Abstract
The Vicinal Risk Minimization principle establishes a bridge between generative models and methods derived from the Structural Risk Minimization Principle such as Support Vector Machines or Statistical Regularization. We explain how VRM provides a framework which integrates a number of existing algorithms, such as Parzen windows, Support Vector Machines, Ridge Regression, Constrained Logistic Classifiers and Tangent-Prop. We then show how the approach implies new algorithms for solving problems usually associated with generative models. New algorithms are described for dealing with pattern recognition problems with very different pattern distributions and dealing with unlabeled data. Preliminary empirical results are presented.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Feature Selection for SVMs

Weston, J., Mukherjee, S., Chapelle, O., Pontil, M., Poggio, T., Vapnik, V.

In Advances in Neural Information Processing Systems 13, pages: 668-674, (Editors: Leen, T.K. , T.G. Dietterich, V. Tresp), MIT Press, Cambridge, MA, USA, Fourteenth Annual Neural Information Processing Systems Conference (NIPS), April 2001 (inproceedings)

Abstract
We introduce a method of feature selection for Support Vector Machines. The method is based upon finding those features which minimize bounds on the leave-one-out error. This search can be efficiently performed via gradient descent. The resulting algorithms are shown to be superior to some standard feature selection algorithms on both toy data and real-life problems of face recognition, pedestrian detection and analyzing DNA microarray data.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Occam’s Razor

Rasmussen, CE., Ghahramani, Z.

In Advances in Neural Information Processing Systems 13, pages: 294-300, (Editors: Leen, T.K. , T.G. Dietterich, V. Tresp), MIT Press, Cambridge, MA, USA, Fourteenth Annual Neural Information Processing Systems Conference (NIPS), April 2001 (inproceedings)

Abstract
The Bayesian paradigm apparently only sometimes gives rise to Occam's Razor; at other times very large models perform well. We give simple examples of both kinds of behaviour. The two views are reconciled when measuring complexity of functions, rather than of the machinery used to implement them. We analyze the complexity of functions for some linear in the parameter models that are equivalent to Gaussian Processes, and always find Occam's Razor at work.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Plaid maskers revisited: asymmetric plaids

Wichmann, F.

pages: 57, 4. T{\"u}binger Wahrnehmungskonferenz (TWK), March 2001 (poster)

Abstract
A large number of psychophysical and physiological experiments suggest that luminance patterns are independently analysed in channels responding to different bands of spatial frequency. There are, however, interactions among stimuli falling well outside the usual estimates of channels' bandwidths. Derrington & Henning (1989) first reported that, in 2-AFC sinusoidal-grating detection, plaid maskers, whose components are oriented symmetrically about the signal orientation, cause a substantially larger threshold elevation than would be predicted from their sinusoidal constituents alone. Wichmann & Tollin (1997a,b) and Wichmann & Henning (1998) confirmed and extended the original findings, measuring masking as a function of presentation time and plaid mask contrast. Here I investigate masking using plaid patterns whose components are asymmetrically positioned about the signal orientation. Standard temporal 2-AFC pattern discrimination experiments were conducted using plaid patterns and oblique sinusoidal gratings as maskers, and horizontally orientated sinusoidal gratings as signals. Signal and maskers were always interleaved on the display (refresh rate 152 Hz). As in the case of the symmetrical plaid maskers, substantial masking was observed for many of the asymmetrical plaids. Masking is neither a straightforward function of the plaid's constituent sinusoidal components nor of the periodicity of the luminance beats between components. These results cause problems for the notion that, even for simple stimuli, detection and discrimination are based on the outputs of channels tuned to limited ranges of spatial frequency and orientation, even if a limited set of nonlinear interactions between these channels is allowed.

Web [BibTex]

Web [BibTex]


no image
Pattern Selection Using the Bias and Variance of Ensemble

Shin, H., Cho, S.

Journal of the Korean Institute of Industrial Engineers, 28(1):112-127, March 2001 (article)

Abstract
[Abstract]: A useful pattern is a pattern that contributes much to learning. For a classification problem those patterns near the class boundary surfaces carry more information to the classifier. For a regression problem the ones near the estimated surface carry more information. In both cases, the usefulness is defined only for those patterns either without error or with negligible error. Using only the useful patterns gives several benefits. First, computational complexity in memory and time for learning is decreased. Second, overfitting is avoided even when the learner is over-sized. Third, learning results in more stable learners. In this paper, we propose a pattern “utility index” that measures the utility of an individual pattern. The utility index is based on the bias and variance of a pattern trained by a network ensemble. In classification, the pattern with a low bias and a high variance gets a high score. In regression, on the other hand, the one with a low bias and a low variance gets a high score. Based on the distribution of the utility index, the original training set is divided into a high-score group and a low-score group. Only the high-score group is then used for training. The proposed method is tested on synthetic and real-world benchmark datasets. The proposed approach gives a better or at least similar performance.

[BibTex]

[BibTex]