Header logo is ei


2011


no image
Combining computational modeling with sparse and low-resolution data

Habeck, M., Nilges, M.

Journal of Structural Biology, 173(3):419, March 2011 (article)

Abstract
Structural biology is moving into a new era by shifting its focus from static structures of single proteins and protein domains to large and often fragile multi-component complexes. Over the past decade, structural genomics initiatives aimed to fill the voids in fold space and to provide a census of all protein structures. Completion of such an atlas of protein structures is still ongoing, but not sufficient for a mechanistic understanding of how living cells function. One of the great challenges is to bridge the gap between atomic resolution detail and the more fuzzy description of the molecular complexes that govern cellular processes or host–pathogen interactions. We want to move from cartoon-like representations of multi-component complexes to atomic resolution structures. To characterize the structures of the increasingly large and often flexible complexes, high resolution structure determination (as was possible for example for the ribosome) will likely stay the exception. Rather, data from many different methods providing information on the shape (X-ray crystallography, electron microscopy, SAXS, AFM, etc.) or on contacts between components (mass spectrometry, co-purification, or spectroscopic methods) need to be integrated with prior structural knowledge to build a consistent model of the complex. A particular difficulty is that the ratio between the number of conformational degrees of freedom and the number of measurements becomes unfavorable as we work with large complexes: data become increasingly sparse. Structural characterization of large molecular assemblies often involves a loss in resolution as well as in number and quality of data. We are good at solving structures of single proteins, but classical high-resolution structure determination by X-ray crystallography and NMR spectroscopy is often facing its limits as we move to higher molecular mass and increased flexibility. Therefore, structural studies on large complexes rely on new experimental techniques that complement the classical high resolution methods. But also computational approaches are becoming more important when it comes to integrating and analyzing structural information of often heterogeneous nature. Cryoelectron microscopy may serve as an example of how experimental methods can benefit from computation. Low-resolution data from cryo-EM show their true power when combined with modeling and bioinformatics methods such rigid docking and secondary structure hunting. Even in high resolution structure determination, molecular modeling is always necessary to calculate structures from data, to complement the missing information and to evaluate and score the obtained structures. With sparse data, all these three aspects become increasingly difficult, and the quality of the modeling approach becomes more important. With data alone, algorithms may not converge any more; scoring against data becomes meaningless; and the potential energy function becomes central not only as a help in making algorithms converge but also to score and evaluate the structures. In addition to the sparsity of the data, hybrid approaches bring the additional difficulty that the different sources of data may have rather different quality, and may be in the extreme case incompatible with each other. In addition to scoring the structures, modeling should also score in some way the data going into the calculation. This special issue brings together some of the numerous efforts to solve the problems that come from sparsity of data and from integrating data from different sources in hybrid approaches. The methods range from predominantly force-field based to mostly data based. Systems of very different sizes, ranging from single domains to multi-component complexes, are treated. We hope that you will enjoy reading the issue and find it a useful and inspiring resource.

PDF DOI [BibTex]

2011

PDF DOI [BibTex]


no image
Batch-Mode Active-Learning Methods for the Interactive Classification of Remote Sensing Images

Demir, B., Persello, C., Bruzzone, L.

IEEE Transactions on Geoscience and Remote Sensing, 49(3):1014-1031, March 2011 (article)

Abstract
This paper investigates different batch-mode active-learning (AL) techniques for the classification of remote sensing (RS) images with support vector machines. This is done by generalizing to multiclass problem techniques defined for binary classifiers. The investigated techniques exploit different query functions, which are based on the evaluation of two criteria: uncertainty and diversity. The uncertainty criterion is associated to the confidence of the supervised algorithm in correctly classifying the considered sample, while the diversity criterion aims at selecting a set of unlabeled samples that are as more diverse (distant one another) as possible, thus reducing the redundancy among the selected samples. The combination of the two criteria results in the selection of the potentially most informative set of samples at each iteration of the AL process. Moreover, we propose a novel query function that is based on a kernel-clustering technique for assessing the diversity of samples and a new strategy for selecting the most informative representative sample from each cluster. The investigated and proposed techniques are theoretically and experimentally compared with state-of-the-art methods adopted for RS applications. This is accomplished by considering very high resolution multispectral and hyperspectral images. By this comparison, we observed that the proposed method resulted in better accuracy with respect to other investigated and state-of-the art methods on both the considered data sets. Furthermore, we derived some guidelines on the design of AL systems for the classification of different types of RS images.

Web DOI [BibTex]

Web DOI [BibTex]


no image
Statistical mechanics analysis of sparse data

Habeck, M.

Journal of Structural Biology, 173(3):541-548, March 2011 (article)

Abstract
Inferential structure determination uses Bayesian theory to combine experimental data with prior structural knowledge into a posterior probability distribution over protein conformational space. The posterior distribution encodes everything one can say objectively about the native structure in the light of the available data and additional prior assumptions and can be searched for structural representatives. Here an analogy is drawn between the posterior distribution and the canonical ensemble of statistical physics. A statistical mechanics analysis assesses the complexity of a structure calculation globally in terms of ensemble properties. Analogs of the free energy and density of states are introduced; partition functions evaluate the consistency of prior assumptions with data. Critical behavior is observed with dwindling restraint density, which impairs structure determination with too sparse data. However, prior distributions with improved realism ameliorate the situation by lowering the critical number of observations. An in-depth analysis of various experimentally accessible structural parameters and force field terms will facilitate a statistical approach to protein structure determination with sparse data that avoids bias as much as possible.

PDF DOI [BibTex]

PDF DOI [BibTex]


no image
Large Scale Bayesian Inference and Experimental Design for Sparse Linear Models

Seeger, M., Nickisch, H.

SIAM Journal on Imaging Sciences, 4(1):166-199, March 2011 (article)

Abstract
Many problems of low-level computer vision and image processing, such as denoising, deconvolution, tomographic reconstruction or super-resolution, can be addressed by maximizing the posterior distribution of a sparse linear model (SLM). We show how higher-order Bayesian decision-making problems, such as optimizing image acquisition in magnetic resonance scanners, can be addressed by querying the SLM posterior covariance, unrelated to the density‘s mode. We propose a scalable algorithmic framework, with which SLM posteriors over full, high-resolution images can be approximated for the first time, solving a variational optimization problem which is convex iff posterior mode finding is convex. These methods successfully drive the optimization of sampling trajectories for real-world magnetic resonance imaging through Bayesian experimental design, which has not been attempted before. Our methodology provides new insight into similarities and differences between sparse reconstruction and approximate Bayesian inference, and has important implications for compressive sensing of real-world images.

Web DOI [BibTex]


no image
Cooperative Cuts

Jegelka, S.

COSA Workshop: Combinatorial Optimization, Statistics, and Applications, March 2011 (talk)

Abstract
Combinatorial problems with submodular cost functions have recently drawn interest. In a standard combinatorial problem, the sum-of-weights cost is replaced by a submodular set function. The result is a powerful model that is though very hard. In this talk, I will introduce cooperative cuts, minimum cuts with submodular edge weights. I will outline methods to approximately solve this problem, and show an application in computer vision. If time permits, the talk will also sketch regret-minimizing online algorithms for submodular-cost combinatorial problems. This is joint work with Jeff Bilmes (University of Washington).

Web [BibTex]

Web [BibTex]


no image
Learning grasp affordance densities

Detry, R., Kraft, D., Kroemer, O., Peters, J., Krüger, N., Piater, J.

Paladyn: Journal of Behavioral Robotics, 2(1):1-17, March 2011 (article)

Abstract
We address the issue of learning and representing object grasp affordance models. We model grasp affordances with continuous probability density functions (grasp densities) which link object-relative grasp poses to their success probability. The underlying function representation is nonparametric and relies on kernel density estimation to provide a continuous model. Grasp densities are learned and refined from exploration, by letting a robot “play” with an object in a sequence of grasp-and-drop actions: the robot uses visual cues to generate a set of grasp hypotheses, which it then executes and records their outcomes. When a satisfactory amount of grasp data is available, an importance-sampling algorithm turns it into a grasp density. We evaluate our method in a largely autonomous learning experiment, run on three objects with distinct shapes. The experiment shows how learning increases success rates. It also measures the success rate of grasps chosen to maximize the probability of success, given reaching constraints.

PDF DOI [BibTex]

PDF DOI [BibTex]


no image
Multiple Kernel Learning: A Unifying Probabilistic Viewpoint

Nickisch, H., Seeger, M.

Max Planck Institute for Biological Cybernetics, March 2011 (techreport)

Abstract
We present a probabilistic viewpoint to multiple kernel learning unifying well-known regularised risk approaches and recent advances in approximate Bayesian inference relaxations. The framework proposes a general objective function suitable for regression, robust regression and classification that is lower bound of the marginal likelihood and contains many regularised risk approaches as special cases. Furthermore, we derive an efficient and provably convergent optimisation algorithm.

Web [BibTex]

Web [BibTex]


no image
Client–Server Multitask Learning From Distributed Datasets

Dinuzzo, F., Pillonetto, G., De Nicolao, G.

IEEE Transactions on Neural Networks, 22(2):290-303, February 2011 (article)

Abstract
A client-server architecture to simultaneously solve multiple learning tasks from distributed datasets is described. In such architecture, each client corresponds to an individual learning task and the associated dataset of examples. The goal of the architecture is to perform information fusion from multiple datasets while preserving privacy of individual data. The role of the server is to collect data in real time from the clients and codify the information in a common database. Such information can be used by all the clients to solve their individual learning task, so that each client can exploit the information content of all the datasets without actually having access to private data of others. The proposed algorithmic framework, based on regularization and kernel methods, uses a suitable class of “mixed effect” kernels. The methodology is illustrated through a simulated recommendation system, as well as an experiment involving pharmacological data coming from a multicentric clinical trial.

DOI [BibTex]

DOI [BibTex]


no image
Extraction of functional information from ongoing brain electrical activity: Extraction en temps-réel d’informations fonctionnelles à partir de l’activité électrique cérébrale

Besserve, M., Martinerie, J.

IRBM, 32(1):27-34, February 2011 (article)

Abstract
The modern analysis of multivariate electrical brain signals requires advanced statistical tools to automatically extract and quantify their information content. These tools include machine learning techniques and information theory. They are currently used both in basic neuroscience and challenging applications such as brain computer interfaces. We review here how these methods have been used at the Laboratoire d’Électroencéphalographie et de Neurophysiologie Appliquée (LENA) to develop a general tool for the real time analysis of functional brain signals. We then give some perspectives on how these tools can help understanding the biological mechanisms of information processing.

PDF DOI [BibTex]


no image
Learning Visual Representations for Perception-Action Systems

Piater, J., Jodogne, S., Detry, R., Kraft, D., Krüger, N., Kroemer, O., Peters, J.

International Journal of Robotics Research, 30(3):294-307, February 2011 (article)

Abstract
We discuss vision as a sensory modality for systems that interact flexibly with uncontrolled environments. Instead of trying to build a generic vision system that produces task-independent representations, we argue in favor of task-specific, learnable representations. This concept is illustrated by two examples of our own work. First, our RLVC algorithm performs reinforcement learning directly on the visual input space. To make this very large space manageable, RLVC interleaves the reinforcement learner with a supervised classification algorithm that seeks to split perceptual states so as to reduce perceptual aliasing. This results in an adaptive discretization of the perceptual space based on the presence or absence of visual features. Its extension, RLJC, additionally handles continuous action spaces. In contrast to the minimalistic visual representations produced by RLVC and RLJC, our second method learns structural object models for robust object detection and pose estimation by probabilistic inference. To these models, the method associates grasp experiences autonomously learned by trial and error. These experiences form a non-parametric representation of grasp success likelihoods over gripper poses, which we call a grasp density. Thus, object detection in a novel scene simultaneously produces suitable grasping options.

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Multi-way set enumeration in weight tensors

Georgii, E., Tsuda, K., Schölkopf, B.

Machine Learning, 82(2):123-155, February 2011 (article)

Abstract
The analysis of n-ary relations receives attention in many different fields, for instance biology, web mining, and social studies. In the basic setting, there are n sets of instances, and each observation associates n instances, one from each set. A common approach to explore these n-way data is the search for n-set patterns, the n-way equivalent of itemsets. More precisely, an n-set pattern consists of specific subsets of the n instance sets such that all possible associations between the corresponding instances are observed in the data. In contrast, traditional itemset mining approaches consider only two-way data, namely items versus transactions. The n-set patterns provide a higher-level view of the data, revealing associative relationships between groups of instances. Here, we generalize this approach in two respects. First, we tolerate missing observations to a certain degree, that means we are also interested in n-sets where most (although not all) of the possible associations have been recorded in the data. Second, we take association weights into account. In fact, we propose a method to enumerate all n-sets that satisfy a minimum threshold with respect to the average association weight. Technically, we solve the enumeration task using a reverse search strategy, which allows for effective pruning of the search space. In addition, our algorithm provides a ranking of the solutions and can consider further constraints. We show experimental results on artificial and real-world datasets from different domains.

PDF DOI [BibTex]

PDF DOI [BibTex]


no image
A graphical model framework for decoding in the visual ERP-based BCI speller

Martens, S., Mooij, J., Hill, N., Farquhar, J., Schölkopf, B.

Neural Computation, 23(1):160-182, January 2011 (article)

Abstract
We present a graphical model framework for decoding in the visual ERP-based speller system. The proposed framework allows researchers to build generative models from which the decoding rules are obtained in a straightforward manner. We suggest two models for generating brain signals conditioned on the stimulus events. Both models incorporate letter frequency information but assume different dependencies between brain signals and stimulus events. For both models, we derive decoding rules and perform a discriminative training. We show on real visual speller data how decoding performance improves by incorporating letter frequency information and using a more realistic graphical model for the dependencies between the brain signals and the stimulus events. Furthermore, we discuss how the standard approach to decoding can be seen as a special case of the graphical model framework. The letter also gives more insight into the discriminative approach for decoding in the visual speller system.

Web DOI [BibTex]

Web DOI [BibTex]


no image
Robust Control of Teleoperation Systems Interacting with Viscoelastic Soft Tissues

Cho, JH., Son, HI., Bhattacharjee, T., Lee, DG., Lee, DY.

IEEE Transactions on Control Systems Technology, January 2011 (article) In revision

[BibTex]

[BibTex]


no image
Effect of Control Parameters and Haptic Cues on Human Perception for Remote Operations

Son, HI., Bhattacharjee, T., Jung, H., Lee, DY.

Experimental Brain Research, January 2011 (article) Submitted

[BibTex]

[BibTex]


no image
Joint Genetic Analysis of Gene Expression Data with Inferred Cellular Phenotypes

Parts, L., Stegle, O., Winn, J., Durbin, R.

PLoS Genetics, 7(1):1-10, January 2011 (article)

Abstract
Even within a defined cell type, the expression level of a gene differs in individual samples. The effects of genotype, measured factors such as environmental conditions, and their interactions have been explored in recent studies. Methods have also been developed to identify unmeasured intermediate factors that coherently influence transcript levels of multiple genes. Here, we show how to bring these two approaches together and analyse genetic effects in the context of inferred determinants of gene expression. We use a sparse factor analysis model to infer hidden factors, which we treat as intermediate cellular phenotypes that in turn affect gene expression in a yeast dataset. We find that the inferred phenotypes are associated with locus genotypes and environmental conditions and can explain genetic associations to genes in trans. For the first time, we consider and find interactions between genotype and intermediate phenotypes inferred from gene expression levels, complementing and extending established results.

Web DOI [BibTex]

Web DOI [BibTex]


no image
Multiple testing, uncertainty and realistic pictures

Langovoy, M., Wittich, O.

(2011-004), EURANDOM, Technische Universiteit Eindhoven, January 2011 (techreport)

Abstract
We study statistical detection of grayscale objects in noisy images. The object of interest is of unknown shape and has an unknown intensity, that can be varying over the object and can be negative. No boundary shape constraints are imposed on the object, only a weak bulk condition for the object's interior is required. We propose an algorithm that can be used to detect grayscale objects of unknown shapes in the presence of nonparametric noise of unknown level. Our algorithm is based on a nonparametric multiple testing procedure. We establish the limit of applicability of our method via an explicit, closed-form, non-asymptotic and nonparametric consistency bound. This bound is valid for a wide class of nonparametric noise distributions. We achieve this by proving an uncertainty principle for percolation on nite lattices.

PDF [BibTex]

PDF [BibTex]


no image
Reinforcement Learning with Bounded Information Loss

Peters, J., Peters, J., Mülling, K., Altun, Y.

AIP Conference Proceedings, 1305(1):365-372, 2011 (article)

Abstract
Policy search is a successful approach to reinforcement learning. However, policy improvements often result in the loss of information. Hence, it has been marred by premature convergence and implausible solutions. As first suggested in the context of covariant or natural policy gradients, many of these problems may be addressed by constraining the information loss. In this paper, we continue this path of reasoning and suggest two reinforcement learning methods, i.e., a model‐based and a model free algorithm that bound the loss in relative entropy while maximizing their return. The resulting methods differ significantly from previous policy gradient approaches and yields an exact update step. It works well on typical reinforcement learning benchmark problems as well as novel evaluations in robotics. We also show a Bayesian bound motivation of this new approach [8].

Web DOI [BibTex]


no image
Nonconvex proximal splitting: batch and incremental algorithms

Sra, S.

(2), Max Planck Institute for Intelligent Systems, Tübingen, Germany, 2011 (techreport)

Abstract
Within the unmanageably large class of nonconvex optimization, we consider the rich subclass of nonsmooth problems having composite objectives (this includes the extensively studied convex, composite objective problems as a special case). For this subclass, we introduce a powerful, new framework that permits asymptotically non-vanishing perturbations. In particular, we develop perturbation-based batch and incremental (online like) nonconvex proximal splitting algorithms. To our knowledge, this is the rst time that such perturbation-based nonconvex splitting algorithms are being proposed and analyzed. While the main contribution of the paper is the theoretical framework, we complement our results by presenting some empirical results on matrix factorization.

PDF [BibTex]

PDF [BibTex]

2003


no image
Learning Control and Planning from the View of Control Theory and Imitation

Peters, J., Schaal, S.

NIPS Workshop "Planning for the Real World: The promises and challenges of dealing with uncertainty", December 2003 (talk)

Abstract
Learning control and planning in high dimensional continuous state-action systems, e.g., as needed in a humanoid robot, has so far been a domain beyond the applicability of generic planning techniques like reinforcement learning and dynamic programming. This talk describes an approach we have taken in order to enable complex robotics systems to learn to accomplish control tasks. Adaptive learning controllers equipped with statistical learning techniques can be used to learn tracking controllers -- missing state information and uncertainty in the state estimates are usually addressed by observers or direct adaptive control methods. Imitation learning is used as an ingredient to seed initial control policies whose output is a desired trajectory suitable to accomplish the task at hand. Reinforcement learning with stochastic policy gradients using a natural gradient forms the third component that allows refining the initial control policy until the task is accomplished. In comparison to general learning control, this approach is highly prestructured and thus more domain specific. However, it seems to be a theoretically clean and feasible strategy for control systems of the complexity that we need to address.

Web [BibTex]

2003

Web [BibTex]


no image
Molecular phenotyping of human chondrocyte cell lines T/C-28a2, T/C-28a4, and C-28/I2

Finger, F., Schorle, C., Zien, A., Gebhard, P., Goldring, M., Aigner, T.

Arthritis & Rheumatism, 48(12):3395-3403, December 2003 (article)

[BibTex]

[BibTex]


no image
A Study on Rainfall - Runoff Models for Improving Ensemble Streamflow Prediction: 1. Rainfallrunoff Models Using Artificial Neural Networks

Jeong, D., Kim, Y., Cho, S., Shin, H.

Journal of the Korean Society of Civil Engineers, 23(6B):521-530, December 2003 (article)

Abstract
The previous ESP (Ensemble Streamflow Prediction) studies conducted in Korea reported that the modeling error is a major source of the ESP forecast error in winter and spring (i.e. dry seasons), and thus suggested that improving the rainfall-runoff model would be critical to obtain more accurate probabilistic forecasts with ESP. This study used two types of Artificial Neural Networks (ANN), such as a Single Neural Network (SNN) and an Ensemble Neural Networks (ENN), to improve the simulation capability of the rainfall-runoff model of the ESP forecasting system for the monthly inflow to the Daecheong dam. Applied for the first time to Korean hydrology, ENN combines the outputs of member models so that it can control the generalization error better than SNN. Because the dry and the flood season in Korea shows considerably different streamflow characteristics, this study calibrated the rainfall-runoff model separately for each season. Therefore, four rainfall-runoff models were developed according to the ANN types and the seasons. This study compared the ANN models with a conceptual rainfall-runoff model called TANK and verified that the ANN models were superior to TANK. Among the ANN models, ENN was more accurate than SNN. The ANN model performance was improved when the model was calibrated separately for the dry and the flood season. The best ANN model developed in this article will be incorporated into the ESP system to increase the forecast capability of ESP for the monthly inflow to the Daecheong dam.

[BibTex]

[BibTex]


no image
Quantitative Cerebral Blood Flow Measurements in the Rat Using a Beta-Probe and H215O

Weber, B., Spaeth, N., Wyss, M., Wild, D., Burger, C., Stanley, R., Buck, A.

Journal of Cerebral Blood Flow and Metabolism, 23(12):1455-1460, December 2003 (article)

Abstract
Beta-probes are a relatively new tool for tracer kinetic studies in animals. They are highly suited to evaluate new positron emission tomography tracers or measure physiologic parameters at rest and after some kind of stimulation or intervention. In many of these experiments, the knowledge of CBF is highly important. Thus, the purpose of this study was to evaluate the method of CBF measurements using a beta-probe and H215O. CBF was measured in the barrel cortex of eight rats at baseline and after acetazolamide challenge. Trigeminal nerve stimulation was additionally performed in five animals. In each category, three injections of 250 to 300 MBq H215O were performed at 10-minute intervals. Data were analyzed using a standard one-tissue compartment model (K1 = CBF, k2 = CBF/p, where p is the partition coefficient). Values for K1 were 0.35 plusminus 0.09, 0.58 plusminus 0.16, and 0.49 plusminus 0.03 mL dot min-1 dot mL-1 at rest, after acetazolamide challenge, and during trigeminal nerve stimulation, respectively. The corresponding values for k2 were 0.55 plusminus 0.12, 0.94 plusminus 0.16, and 0.85 plusminus 0.12 min-7, and for p were 0.64 plusminus 0.05, 0.61 plusminus 0.07, and 0.59 plusminus 0.06.The standard deviation of the difference between two successive experiments, a measure for the reproducibility of the method, was 10.1%, 13.0%, and 5.7% for K1, k2, and p, respectively. In summary, beta-probes in conjunction with H215O allow the reproducible quantitative measurement of CBF, although some systematic underestimation seems to occur, probably because of partial volume effects.

PDF DOI [BibTex]

PDF DOI [BibTex]


no image
Recurrent neural networks from learning attractor dynamics

Schaal, S., Peters, J.

NIPS Workshop on RNNaissance: Recurrent Neural Networks, December 2003 (talk)

Abstract
Many forms of recurrent neural networks can be understood in terms of dynamic systems theory of difference equations or differential equations. Learning in such systems corresponds to adjusting some internal parameters to obtain a desired time evolution of the network, which can usually be characterized in term of point attractor dynamics, limit cycle dynamics, or, in some more rare cases, as strange attractor or chaotic dynamics. Finding a stable learning process to adjust the open parameters of the network towards shaping the desired attractor type and basin of attraction has remain a complex task, as the parameter trajectories during learning can lead the system through a variety of undesirable unstable behaviors, such that learning may never succeed. In this presentation, we review a recently developed learning framework for a class of recurrent neural networks that employs a more structured network approach. We assume that the canonical system behavior is known a priori, e.g., it is a point attractor or a limit cycle. With either supervised learning or reinforcement learning, it is possible to acquire the transformation from a simple representative of this canonical behavior (e.g., a 2nd order linear point attractor, or a simple limit cycle oscillator) to the desired highly complex attractor form. For supervised learning, one shot learning based on locally weighted regression techniques is possible. For reinforcement learning, stochastic policy gradient techniques can be employed. In any case, the recurrent network learned by these methods inherits the stability properties of the simple dynamic system that underlies the nonlinear transformation, such that stability of the learning approach is not a problem. We demonstrate the success of this approach for learning various skills on a humanoid robot, including tasks that require to incorporate additional sensory signals as coupling terms to modify the recurrent network evolution on-line.

Web [BibTex]

Web [BibTex]


no image
Support Vector Channel Selection in BCI

Lal, T., Schröder, M., Hinterberger, T., Weston, J., Bogdan, M., Birbaumer, N., Schölkopf, B.

(120), Max Planck Institute for Biological Cybernetics, Tuebingen, Germany, December 2003 (techreport)

Abstract
Designing a Brain Computer Interface (BCI) system one can choose from a variety of features that may be useful for classifying brain activity during a mental task. For the special case of classifying EEG signals we propose the usage of the state of the art feature selection algorithms Recursive Feature Elimination [3] and Zero-Norm Optimization [13] which are based on the training of Support Vector Machines (SVM) [11]. These algorithms can provide more accurate solutions than standard filter methods for feature selection [14]. We adapt the methods for the purpose of selecting EEG channels. For a motor imagery paradigm we show that the number of used channels can be reduced significantly without increasing the classification error. The resulting best channels agree well with the expected underlying cortical activity patterns during the mental tasks. Furthermore we show how time dependent task specific information can be visualized.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Blind separation of post-nonlinear mixtures using linearizing transformations and temporal decorrelation

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

Journal of Machine Learning Research, 4(7-8):1319-1338, November 2003 (article)

Abstract
We propose two methods that reduce the post-nonlinear blind source separation problem (PNL-BSS) to a linear BSS problem. The first method is based on the concept of maximal correlation: we apply the alternating conditional expectation (ACE) algorithm--a powerful technique from non-parametric statistics--to approximately invert the componentwise nonlinear functions. The second method is a Gaussianizing transformation, which is motivated by the fact that linearly mixed signals before nonlinear transformation are approximately Gaussian distributed. This heuristic, but simple and efficient procedure works as good as the ACE method. Using the framework provided by ACE, convergence can be proven. The optimal transformations obtained by ACE coincide with the sought-after inverse functions of the nonlinearities. After equalizing the nonlinearities, temporal decorrelation separation (TDSEP) allows us to recover the source signals. Numerical simulations testing "ACE-TD" and "Gauss-TD" on realistic examples are performed with excellent results.

PDF PDF DOI [BibTex]

PDF PDF DOI [BibTex]


no image
Correlated stage- and subfield-associated hippocampal gene expression patterns in experimental and human temporal lobe epilepsy

Becker, A., Chen, J., Zien, A., Sochivko, D., Normann, S., Schramm, J., Elger, C., Wiestler, O., Blumcke, I.

European Journal of Neuroscience, 18(10):2792-2802, November 2003 (article)

Abstract
Epileptic activity evokes profound alterations of hippocampal organization and function. Genomic responses may reflect immediate consequences of excitatory stimulation as well as sustained molecular processes related to neuronal plasticity and structural remodeling. Using oligonucleotide microarrays with 8799 sequences, we determined subregional gene expression profiles in rats subjected to pilocarpine-induced epilepsy (U34A arrays, Affymetrix, Santa Clara, CA, USA; P < 0.05, twofold change, n = 3 per stage). Patterns of gene expression corresponded to distinct stages of epilepsy development. The highest number of differentially expressed genes (dentate gyrus, approx. 400 genes and CA1, approx. 700 genes) was observed 3 days after status epilepticus. The majority of up-regulated genes was associated with mechanisms of cellular stress and injury - 14 days after status epilepticus, numerous transcription factors and genes linked to cytoskeletal and synaptic reorganization were differentially expressed and, in the stage of chronic spontaneous seizures, distinct changes were observed in the transcription of genes involved in various neurotransmission pathways and between animals with low vs. high seizure frequency. A number of genes (n = 18) differentially expressed during the chronic epileptic stage showed corresponding expression patterns in hippocampal subfields of patients with pharmacoresistant temporal lobe epilepsy (n = 5 temporal lobe epilepsy patients; U133A microarrays, Affymetrix; covering 22284 human sequences). These data provide novel insights into the molecular mechanisms of epileptogenesis and seizure-associated cellular and structural remodeling of the hippocampus.

[BibTex]

[BibTex]


no image
Concentration Inequalities for Sub-Additive Functions Using the Entropy Method

Bousquet, O.

Stochastic Inequalities and Applications, 56, pages: 213-247, Progress in Probability, (Editors: Giné, E., C. Houdré and D. Nualart), November 2003 (article)

Abstract
We obtain exponential concentration inequalities for sub-additive functions of independent random variables under weak conditions on the increments of those functions, like the existence of exponential moments for these increments. As a consequence of these general inequalities, we obtain refinements of Talagrand's inequality for empirical processes and new bounds for randomized empirical processes. These results are obtained by further developing the entropy method introduced by Ledoux.

PostScript [BibTex]

PostScript [BibTex]


no image
Technical report on Separation methods for nonlinear mixtures

Jutten, C., Karhunen, J., Almeida, L., Harmeling, S.

(D29), EU-Project BLISS, October 2003 (techreport)

PDF [BibTex]

PDF [BibTex]


no image
Image Reconstruction by Linear Programming

Tsuda, K., Rätsch, G.

(118), Max Planck Institute for Biological Cybernetics, Tübingen, Germany, October 2003 (techreport)

PDF [BibTex]

PDF [BibTex]


no image
YKL-39 (chitinase 3-like protein 2), but not YKL-40 (chitinase 3-like protein 1), is up regulated in osteoarthritic chondrocytes

Knorr, T., Obermayr, F., Bartnik, E., Zien, A., Aigner, T.

Annals of the Rheumatic Diseases, 62(10):995-998, October 2003 (article)

Abstract
OBJECTIVE: To investigate quantitatively the mRNA expression levels of YKL-40, an established marker of rheumatoid and osteoarthritic cartilage degeneration in synovial fluid and serum, and a closely related molecule YKL-39, in articular chondrocytes. METHODS: cDNA array and online quantitative polymerase chain reaction (PCR) were used to measure mRNA expression levels of YKL-39 and YKL-40 in chondrocytes in normal, early degenerative, and late stage osteoarthritic cartilage samples. RESULTS: Expression analysis showed high levels of both proteins in normal articular chondrocytes, with lower levels of YKL-39 than YKL-40. Whereas YKL-40 was significantly down regulated in late stage osteoarthritic chondrocytes, YKL-39 was significantly up regulated. In vitro both YKLs were down regulated by interleukin 1beta. CONCLUSIONS: The up regulation of YKL-39 in osteoarthritic cartilage suggests that YKL-39 may be a more accurate marker of chondrocyte activation than YKL-40, although it has yet to be established as a suitable marker in synovial fluid and serum. The decreased expression of YKL-40 by osteoarthritic chondrocytes is surprising as increased levels have been reported in rheumatoid and osteoarthritic synovial fluid, where it may derive from activated synovial cells or osteophytic tissue or by increased matrix destruction in the osteoarthritic joint. YKL-39 and YKL-40 are potentially interesting marker molecules for arthritic joint disease because they are abundantly expressed by both normal and osteoarthritic chondrocytes.

[BibTex]

[BibTex]


no image
Technical report on implementation of linear methods and validation on acoustic sources

Harmeling, S., Bünau, P., Ziehe, A., Pham, D.

EU-Project BLISS, September 2003 (techreport)

PDF [BibTex]

PDF [BibTex]


no image
Statistical Learning Theory

Bousquet, O.

Machine Learning Summer School, August 2003 (talk)

PDF [BibTex]

PDF [BibTex]


no image
Remarks on Statistical Learning Theory

Bousquet, O.

Machine Learning Summer School, August 2003 (talk)

PDF [BibTex]

PDF [BibTex]


no image
Statistical Learning Theory, Capacity and Complexity

Schölkopf, B.

Complexity, 8(4):87-94, July 2003 (article)

Abstract
We give an exposition of the ideas of statistical learning theory, followed by a discussion of how a reinterpretation of the insights of learning theory could potentially also benefit our understanding of a certain notion of complexity.

Web DOI [BibTex]


no image
Ranking on Data Manifolds

Zhou, D., Weston, J., Gretton, A., Bousquet, O., Schölkopf, B.

(113), Max Planck Institute for Biological Cybernetics, 72076 Tuebingen, Germany, June 2003 (techreport)

Abstract
The Google search engine has had a huge success with its PageRank web page ranking algorithm, which exploits global, rather than local, hyperlink structure of the World Wide Web using random walk. This algorithm can only be used for graph data, however. Here we propose a simple universal ranking algorithm for vectorial data, based on the exploration of the intrinsic global geometric structure revealed by a huge amount of data. Experimental results from image and text to bioinformatics illustrates the validity of our algorithm.

PDF [BibTex]

PDF [BibTex]


no image
Kernel Hebbian Algorithm for Iterative Kernel Principal Component Analysis

Kim, K., Franz, M., Schölkopf, B.

(109), MPI f. biologische Kybernetik, Tuebingen, June 2003 (techreport)

Abstract
A new method for performing a kernel principal component analysis is proposed. By kernelizing the generalized Hebbian algorithm, one can iteratively estimate the principal components in a reproducing kernel Hilbert space with only linear order memory complexity. The derivation of the method, a convergence proof, and preliminary applications in image hyperresolution are presented. In addition, we discuss the extension of the method to the online learning of kernel principal components.

PDF [BibTex]

PDF [BibTex]


no image
Learning with Local and Global Consistency

Zhou, D., Bousquet, O., Lal, T., Weston, J., Schölkopf, B.

(112), Max Planck Institute for Biological Cybernetics, Tuebingen, Germany, June 2003 (techreport)

Abstract
We consider the learning problem in the transductive setting. Given a set of points of which only some are labeled, the goal is to predict the label of the unlabeled points. A principled clue to solve such a learning problem is the consistency assumption that a classifying function should be sufficiently smooth with respect to the structure revealed by these known labeled and unlabeled points. We present a simple algorithm to obtain such a smooth solution. Our method yields encouraging experimental results on a number of classification problems and demonstrates effective use of unlabeled data.

[BibTex]

[BibTex]


no image
Dealing with large Diagonals in Kernel Matrices

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

Annals of the Institute of Statistical Mathematics, 55(2):391-408, June 2003 (article)

Abstract
In kernel methods, all the information about the training data is contained in the Gram matrix. If this matrix has large diagonal values, which arises for many types of kernels, then kernel methods do not perform well: We propose and test several methods for dealing with this problem by reducing the dynamic range of the matrix while preserving the positive definiteness of the Hessian of the quadratic programming problem that one has to solve when training a Support Vector Machine, which is a common kernel approach for pattern recognition.

PDF DOI [BibTex]

PDF DOI [BibTex]


no image
The Metric Nearness Problem with Applications

Dhillon, I., Sra, S., Tropp, J.

Univ. of Texas at Austin, June 2003 (techreport)

GZIP [BibTex]

GZIP [BibTex]


no image
Implicit Wiener Series

Franz, M., Schölkopf, B.

(114), Max Planck Institute for Biological Cybernetics, June 2003 (techreport)

Abstract
The Wiener series is one of the standard methods to systematically characterize the nonlinearity of a neural system. The classical estimation method of the expansion coefficients via cross-correlation suffers from severe problems that prevent its application to high-dimensional and strongly nonlinear systems. We propose a new estimation method based on regression in a reproducing kernel Hilbert space that overcomes these problems. Numerical experiments show performance advantages in terms of convergence, interpretability and system size that can be handled.

PDF [BibTex]

PDF [BibTex]


no image
Machine Learning approaches to protein ranking: discriminative, semi-supervised, scalable algorithms

Weston, J., Leslie, C., Elisseeff, A., Noble, W.

(111), Max Planck Institute for Biological Cybernetics, Tübingen, Germany, June 2003 (techreport)

Abstract
A key tool in protein function discovery is the ability to rank databases of proteins given a query amino acid sequence. The most successful method so far is a web-based tool called PSI-BLAST which uses heuristic alignment of a profile built using the large unlabeled database. It has been shown that such use of global information via an unlabeled data improves over a local measure derived from a basic pairwise alignment such as performed by PSI-BLAST's predecessor, BLAST. In this article we look at ways of leveraging techniques from the field of machine learning for the problem of ranking. We show how clustering and semi-supervised learning techniques, which aim to capture global structure in data, can significantly improve over PSI-BLAST.

PDF [BibTex]

PDF [BibTex]


no image
The em Algorithm for Kernel Matrix Completion with Auxiliary Data

Tsuda, K., Akaho, S., Asai, K.

Journal of Machine Learning Research, 4, pages: 67-81, May 2003 (article)

PDF [BibTex]

PDF [BibTex]


no image
Constructing Descriptive and Discriminative Non-linear Features: Rayleigh Coefficients in Kernel Feature Spaces

Mika, S., Rätsch, G., Weston, J., Schölkopf, B., Smola, A., Müller, K.

IEEE Transactions on Pattern Analysis and Machine Intelligence, 25(5):623-628, May 2003 (article)

Abstract
We incorporate prior knowledge to construct nonlinear algorithms for invariant feature extraction and discrimination. Employing a unified framework in terms of a nonlinearized variant of the Rayleigh coefficient, we propose nonlinear generalizations of Fisher‘s discriminant and oriented PCA using support vector kernel functions. Extensive simulations show the utility of our approach.

DOI [BibTex]

DOI [BibTex]


no image
The Geometry Of Kernel Canonical Correlation Analysis

Kuss, M., Graepel, T.

(108), Max Planck Institute for Biological Cybernetics, Tübingen, Germany, May 2003 (techreport)

Abstract
Canonical correlation analysis (CCA) is a classical multivariate method concerned with describing linear dependencies between sets of variables. After a short exposition of the linear sample CCA problem and its analytical solution, the article proceeds with a detailed characterization of its geometry. Projection operators are used to illustrate the relations between canonical vectors and variates. The article then addresses the problem of CCA between spaces spanned by objects mapped into kernel feature spaces. An exact solution for this kernel canonical correlation (KCCA) problem is derived from a geometric point of view. It shows that the expansion coefficients of the canonical vectors in their respective feature space can be found by linear CCA in the basis induced by kernel principal component analysis. The effect of mappings into higher dimensional feature spaces is considered critically since it simplifies the CCA problem in general. Then two regularized variants of KCCA are discussed. Relations to other methods are illustrated, e.g., multicategory kernel Fisher discriminant analysis, kernel principal component regression and possible applications thereof in blind source separation.

PDF [BibTex]

PDF [BibTex]


no image
Kernel-based nonlinear blind source separation

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

Neural Computation, 15(5):1089-1124, May 2003 (article)

Abstract
We propose kTDSEP, a kernel-based algorithm for nonlinear blind source separation (BSS). It combines complementary research fields: kernel feature spaces and BSS using temporal information. This yields an efficient algorithm for nonlinear BSS with invertible nonlinearity. Key assumptions are that the kernel feature space is chosen rich enough to approximate the nonlinearity and that signals of interest contain temporal information. Both assumptions are fulfilled for a wide set of real-world applications. The algorithm works as follows: First, the data are (implicitly) mapped to a high (possibly infinite)—dimensional kernel feature space. In practice, however, the data form a smaller submanifold in feature space—even smaller than the number of training data points—a fact that has already been used by, for example, reduced set techniques for support vector machines. We propose to adapt to this effective dimension as a preprocessing step and to construct an orthonormal basis of this submanifold. The latter dimension-reduction step is essential for making the subsequent application of BSS methods computationally and numerically tractable. In the reduced space, we use a BSS algorithm that is based on second-order temporal decorrelation. Finally, we propose a selection procedure to obtain the original sources from the extracted nonlinear components automatically. Experiments demonstrate the excellent performance and efficiency of our kTDSEP algorithm for several problems of nonlinear BSS and for more than two sources.

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
The Kernel Mutual Information

Gretton, A., Herbrich, R., Smola, A.

Max Planck Institute for Biological Cybernetics, April 2003 (techreport)

Abstract
We introduce two new functions, the kernel covariance (KC) and the kernel mutual information (KMI), to measure the degree of independence of several continuous random variables. The former is guaranteed to be zero if and only if the random variables are pairwise independent; the latter shares this property, and is in addition an approximate upper bound on the mutual information, as measured near independence, and is based on a kernel density estimate. We show that Bach and Jordan‘s kernel generalised variance (KGV) is also an upper bound on the same kernel density estimate, but is looser. Finally, we suggest that the addition of a regularising term in the KGV causes it to approach the KMI, which motivates the introduction of this regularisation. The performance of the KC and KMI is verified in the context of instantaneous independent component analysis (ICA), by recovering both artificial and real (musical) signals following linear mixing.

PostScript [BibTex]

PostScript [BibTex]


no image
Tractable Inference for Probabilistic Data Models

Csato, L., Opper, M., Winther, O.

Complexity, 8(4):64-68, April 2003 (article)

Abstract
We present an approximation technique for probabilistic data models with a large number of hidden variables, based on ideas from statistical physics. We give examples for two nontrivial applications. © 2003 Wiley Periodicals, Inc.

PDF GZIP Web [BibTex]

PDF GZIP Web [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.

Bioinformatics, 19(6):764-771, April 2003 (article)

Abstract
Motivation: In drug discovery a key task is to identify characteristics that separate active (binding) compounds from inactive (non-binding) ones. An automated prediction system can help reduce resources necessary to carry out this task. Results: Two methods for prediction of molecular bioactivity for drug design are introduced and shown to perform well in a data set previously studied as part of the KDD (Knowledge Discovery and Data Mining) Cup 2001. The data is characterized by very few positive examples, a very large number of features (describing three-dimensional properties of the molecules) and rather different distributions between training and test data. Two techniques are introduced specifically to tackle these problems: a feature selection method for unbalanced data and a classifier which adapts to the distribution of the the unlabeled test data (a so-called transductive method). We show both techniques improve identification performance and in conjunction provide an improvement over using only one of the techniques. Our results suggest the importance of taking into account the characteristics in this data which may also be relevant in other problems of a similar type.

Web [BibTex]


no image
Rademacher and Gaussian averages in Learning Theory

Bousquet, O.

Universite de Marne-la-Vallee, March 2003 (talk)

PDF [BibTex]

PDF [BibTex]