Header logo is


2008


no image
Learning to Localize Objects with Structured Output Regression

Blaschko, MB., Lampert, CH.

In ECCV 2008, pages: 2-15, (Editors: Forsyth, D. A., P. H.S. Torr, A. Zisserman), Springer, Berlin, Germany, 10th European Conference on Computer Vision, October 2008, Best Student Paper Award (inproceedings)

Abstract
Sliding window classifiers are among the most successful and widely applied techniques for object localization. However, training is typically done in a way that is not specific to the localization task. First a binary classifier is trained using a sample of positive and negative examples, and this classifier is subsequently applied to multiple regions within test images. We propose instead to treat object localization in a principled way by posing it as a problem of predicting structured data: we model the problem not as binary classification, but as the prediction of the bounding box of objects located in images. The use of a joint-kernel framework allows us to formulate the training procedure as a generalization of an SVM, which can be solved efficiently. We further improve computational efficiency by using a branch-and-bound strategy for localization during both training and testing. Experimental evaluation on the PASCAL VOC and TU Darmstadt datasets show that the structured training procedure improves pe rformance over binary training as well as the best previously published scores.

ei

PDF Web DOI [BibTex]

2008


PDF Web DOI [BibTex]


no image
Automatic Image Colorization Via Multimodal Predictions

Charpiat, G., Hofmann, M., Schölkopf, B.

In Computer Vision - ECCV 2008, Lecture Notes in Computer Science, Vol. 5304, pages: 126-139, (Editors: DA Forsyth and PHS Torr and A Zisserman), Springer, Berlin, Germany, 10th European Conference on Computer Vision, October 2008 (inproceedings)

Abstract
We aim to color automatically greyscale images, without any manual intervention. The color proposition could then be interactively corrected by user-provided color landmarks if necessary. Automatic colorization is nontrivial since there is usually no one-to-one correspondence between color and local texture. The contribution of our framework is that we deal directly with multimodality and estimate, for each pixel of the image to be colored, the probability distribution of all possible colors, instead of choosing the most probable color at the local level. We also predict the expected variation of color at each pixel, thus defining a nonuniform spatial coherency criterion. We then use graph cuts to maximize the probability of the whole colored image at the global level. We work in the L-a-b color space in order to approximate the human perception of distances between colors, and we use machine learning tools to extract as much information as possible from a dataset of colored examples. The resulting algorithm is fast, designed to be more robust to texture noise, and is above all able to deal with ambiguity, in contrary to previous approaches.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Nonparametric Independence Tests: Space Partitioning and Kernel Approaches

Gretton, A., Györfi, L.

In ALT08, pages: 183-198, (Editors: Freund, Y. , L. Györfi, G. Turán, T. Zeugmann), Springer, Berlin, Germany, 19th International Conference on Algorithmic Learning Theory (ALT08), October 2008 (inproceedings)

Abstract
Three simple and explicit procedures for testing the independence of two multi-dimensional random variables are described. Two of the associated test statistics (L1, log-likelihood) are defined when the empirical distribution of the variables is restricted to finite partitions. A third test statistic is defined as a kernel-based independence measure. All tests reject the null hypothesis of independence if the test statistics become large. The large deviation and limit distribution properties of all three test statistics are given. Following from these results, distributionfree strong consistent tests of independence are derived, as are asymptotically alpha-level tests. The performance of the tests is evaluated experimentally on benchmark data.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Automatic 3D Face Reconstruction from Single Images or Video

Breuer, P., Kim, K., Kienzle, W., Schölkopf, B., Blanz, V.

In FG 2008, pages: 1-8, IEEE Computer Society, Los Alamitos, CA, USA, 8th IEEE International Conference on Automatic Face and Gesture Recognition, September 2008 (inproceedings)

Abstract
This paper presents a fully automated algorithm for reconstructing a textured 3D model of a face from a single photograph or a raw video stream. The algorithm is based on a combination of Support Vector Machines (SVMs) and a Morphable Model of 3D faces. After SVM face detection, individual facial features are detected using a novel regression- and classification-based approach, and probabilistically plausible configurations of features are selected to produce a list of candidates for several facial feature positions. In the next step, the configurations of feature points are evaluated using a novel criterion that is based on a Morphable Model and a combination of linear projections. To make the algorithm robust with respect to head orientation, this process is iterated while the estimate of pose is refined. Finally, the feature points initialize a model-fitting procedure of the Morphable Model. The result is a highresolution 3D surface model.

ei

PDF DOI [BibTex]

PDF DOI [BibTex]


no image
Kernel Measures of Conditional Dependence

Fukumizu, K., Gretton, A., Sun, X., Schölkopf, B.

In Advances in neural information processing systems 20, pages: 489-496, (Editors: JC Platt and D Koller and Y Singer and S Roweis), Curran, Red Hook, NY, USA, 21st Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
We propose a new measure of conditional dependence of random variables, based on normalized cross-covariance operators on reproducing kernel Hilbert spaces. Unlike previous kernel dependence measures, the proposed criterion does not depend on the choice of kernel in the limit of infinite data, for a wide class of kernels. At the same time, it has a straightforward empirical estimate with good convergence behaviour. We discuss the theoretical properties of the measure, and demonstrate its application in experiments.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
An Analysis of Inference with the Universum

Sinz, F., Chapelle, O., Agarwal, A., Schölkopf, B.

In Advances in neural information processing systems 20, pages: 1369-1376, (Editors: JC Platt and D Koller and Y Singer and S Roweis), Curran, Red Hook, NY, USA, 21st Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
We study a pattern classification algorithm which has recently been proposed by Vapnik and coworkers. It builds on a new inductive principle which assumes that in addition to positive and negative data, a third class of data is available, termed the Universum. We assay the behavior of the algorithm by establishing links with Fisher discriminant analysis and oriented PCA, as well as with an SVM in a projected subspace (or, equivalently, with a data-dependent reduced kernel). We also provide experimental results.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Learning with Transformation Invariant Kernels

Walder, C., Chapelle, O.

In Advances in neural information processing systems 20, pages: 1561-1568, (Editors: Platt, J. C., D. Koller, Y. Singer, S. Roweis), Curran, Red Hook, NY, USA, Twenty-First Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
This paper considers kernels invariant to translation, rotation and dilation. We show that no non-trivial positive definite (p.d.) kernels exist which are radial and dilation invariant, only conditionally positive definite (c.p.d.) ones. Accordingly, we discuss the c.p.d. case and provide some novel analysis, including an elementary derivation of a c.p.d. representer theorem. On the practical side, we give a support vector machine (s.v.m.) algorithm for arbitrary c.p.d. kernels. For the thinplate kernel this leads to a classifier with only one parameter (the amount of regularisation), which we demonstrate to be as effective as an s.v.m. with the Gaussian kernel, even though the Gaussian involves a second parameter (the length scale).

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Episodic Reinforcement Learning by Logistic Reward-Weighted Regression

Wierstra, D., Schaul, T., Peters, J., Schmidhuber, J.

In ICANN 2008, pages: 407-416, (Editors: Kurkova-Pohlova, V. , R. Neruda, J. Koutnik), Springer, Berlin, Germany, 18th International Conference on Artificial Neural Networks, September 2008 (inproceedings)

Abstract
It has been a long-standing goal in the adaptive control community to reduce the generically difficult, general reinforcement learning (RL) problem to simpler problems solvable by supervised learning. While this approach is today’s standard for value function-based methods, fewer approaches are known that apply similar reductions to policy search methods. Recently, it has been shown that immediate RL problems can be solved by reward-weighted regression, and that the resulting algorithm is an expectation maximization (EM) algorithm with strong guarantees. In this paper, we extend this algorithm to the episodic case and show that it can be used in the context of LSTM recurrent neural networks (RNNs). The resulting RNN training algorithm is equivalent to a weighted self-modeling supervised learning technique. We focus on partially observable Markov decision problems (POMDPs) where it is essential that the policy is nonstationary in order to be optimal. We show that this new reward-weighted logistic regression u sed in conjunction with an RNN architecture can solve standard benchmark POMDPs with ease.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Local Gaussian Processes Regression for Real-time Model-based Robot Control

Nguyen-Tuong, D., Peters, J.

In IROS 2008, pages: 380-385, IEEE Service Center, Piscataway, NJ, USA, 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems, September 2008 (inproceedings)

Abstract
High performance and compliant robot control require accurate dynamics models which cannot be obtained analytically for sufficiently complex robot systems. In such cases, machine learning offers a promising alternative for approximating the robot dynamics using measured data. This approach offers a natural framework to incorporate unknown nonlinearities as well as to continually adapt online for changes in the robot dynamics. However, the most accurate regression methods, e.g. Gaussian processes regression (GPR) and support vector regression (SVR), suffer from exceptional high computational complexity which prevents their usage for large numbers of samples or online learning to date. Inspired by locally linear regression techniques, we propose an approximation to the standard GPR using local Gaussian processes models. Due to reduced computational cost, local Gaussian processes (LGP) can be applied for larger sample-sizes and online learning. Comparisons with other nonparametric regressions, e.g. standard GPR, nu-SVR and locally weighted projection regression (LWPR), show that LGP has higher accuracy than LWPR close to the performance of standard GPR and nu-SVR while being sufficiently fast for online learning.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Colored Maximum Variance Unfolding

Song, L., Smola, A., Borgwardt, K., Gretton, A.

In Advances in neural information processing systems 20, pages: 1385-1392, (Editors: Platt, J. C., D. Koller, Y. Singer, S. Roweis), Curran, Red Hook, NY, USA, Twenty-First Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
Maximum variance unfolding (MVU) is an effective heuristic for dimensionality reduction. It produces a low-dimensional representation of the data by maximizing the variance of their embeddings while preserving the local distances of the original data. We show that MVU also optimizes a statistical dependence measure which aims to retain the identity of individual observations under the distancepreserving constraints. This general view allows us to design "colored" variants of MVU, which produce low-dimensional representations for a given task, e.g. subject to class labels or other side information.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Assessing Nonlinear Granger Causality from Multivariate Time Series

Sun, X.

In Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2008, pages: 440-455, (Editors: Daelemans, W. , B. Goethals, K. Morik), Springer, Berlin, Germany, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD), September 2008 (inproceedings)

Abstract
A straightforward nonlinear extension of Granger’s concept of causality in the kernel framework is suggested. The kernel-based approach to assessing nonlinear Granger causality in multivariate time series enables us to determine, in a model-free way, whether the causal relation between two time series is present or not and whether it is direct or mediated by other processes. The trace norm of the so-called covariance operator in feature space is used to measure the prediction error. Relying on this measure, we test the improvement of predictability between time series by subsampling-based multiple testing. The distributional properties of the resulting p-values reveal the direction of Granger causality. Experiments with simulated and real-world data show that our method provides encouraging results.

ei

PDF PDF DOI [BibTex]

PDF PDF DOI [BibTex]


no image
Near-Maximum Entropy Models for Binary Neural Representations of Natural Images

Bethge, M., Berens, P.

In Advances in neural information processing systems 20, pages: 97-104, (Editors: Platt, J. C., D. Koller, Y. Singer, S. Roweis), Curran, Red Hook, NY, USA, Twenty-First Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
Maximum entropy analysis of binary variables provides an elegant way for studying the role of pairwise correlations in neural populations. Unfortunately, these approaches suffer from their poor scalability to high dimensions. In sensory coding, however, high-dimensional data is ubiquitous. Here, we introduce a new approach using a near-maximum entropy model, that makes this type of analysis feasible for very high-dimensional data - the model parameters can be derived in closed form and sampling is easy. We demonstrate its usefulness by studying a simple neural representation model of natural images. For the first time, we are able to directly compare predictions from a pairwise maximum entropy model not only in small groups of neurons, but also in larger populations of more than thousand units. Our results indicate that in such larger networks interactions exist that are not predicted by pairwise correlations, despite the fact that pairwise correlations explain the lower-dimensional marginal statistics extrem ely well up to the limit of dimensionality where estimation of the full joint distribution is feasible.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Learning Perceptual Coupling for Motor Primitives

Kober, J., Mohler, B., Peters, J.

In IROS 2008, pages: 834-839, IEEE Service Center, Piscataway, NJ, USA, 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems, September 2008 (inproceedings)

Abstract
Dynamic system-based motor primitives [1] have enabled robots to learn complex tasks ranging from Tennisswings to locomotion. However, to date there have been only few extensions which have incorporated perceptual coupling to variables of external focus, and, furthermore, these modifications have relied upon handcrafted solutions. Humans learn how to couple their movement primitives with external variables. Clearly, such a solution is needed in robotics. In this paper, we propose an augmented version of the dynamic systems motor primitives which incorporates perceptual coupling to an external variable. The resulting perceptually driven motor primitives include the previous primitives as a special case and can inherit some of their interesting properties. We show that these motor primitives can perform complex tasks such as Ball-in-a-Cup or Kendama task even with large variances in the initial conditions where a skilled human player would be challenged. For doing so, we initialize the motor primitives in the traditional way by imitation learning without perceptual coupling. Subsequently, we improve the motor primitives using a novel reinforcement learning method which is particularly well-suited for motor primitives.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Receptive Fields without Spike-Triggering

Macke, J., Zeck, G., Bethge, M.

In Advances in neural information processing systems 20, pages: 969-976, (Editors: Platt, J. C., D. Koller, Y. Singer, S. Roweis), Curran, Red Hook, NY, USA, Twenty-First Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
Stimulus selectivity of sensory neurons is often characterized by estimating their receptive field properties such as orientation selectivity. Receptive fields are usually derived from the mean (or covariance) of the spike-triggered stimulus ensemble. This approach treats each spike as an independent message but does not take into account that information might be conveyed through patterns of neural activity that are distributed across space or time. Can we find a concise description for the processing of a whole population of neurons analogous to the receptive field for single neurons? Here, we present a generalization of the linear receptive field which is not bound to be triggered on individual spikes but can be meaningfully linked to distributed response patterns. More precisely, we seek to identify those stimulus features and the corresponding patterns of neural activity that are most reliably coupled. We use an extension of reverse-correlation methods based on canonical correlation analysis. The resulting population receptive fields span the subspace of stimuli that is most informative about the population response. We evaluate our approach using both neuronal models and multi-electrode recordings from rabbit retinal ganglion cells. We show how the model can be extended to capture nonlinear stimulus-response relationships using kernel canonical correlation analysis, which makes it possible to test different coding mechanisms. Our technique can also be used to calculate receptive fields from multi-dimensional neural measurements such as those obtained from dynamic imaging methods.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
An Automated Combination of Kernels for Predicting Protein Subcellular Localization

Ong, C., Zien, A.

In WABI 2008, pages: 186-197, (Editors: Crandall, K. A., J. Lagergren), Springer, Berlin, Germany, 8th Workshop on Algorithms in Bioinformatics, September 2008 (inproceedings)

Abstract
Protein subcellular localization is a crucial ingredient to many important inferences about cellular processes, including prediction of protein function and protein interactions. While many predictive computational tools have been proposed, they tend to have complicated architectures and require many design decisions from the developer. Here we utilize the multiclass support vector machine (m-SVM) method to directly solve protein subcellular localization without resorting to the common approach of splitting the problem into several binary classification problems. We further propose a general class of protein sequence kernels which considers all motifs, including motifs with gaps. Instead of heuristically selecting one or a few kernels from this family, we utilize a recent extension of SVMs that optimizes over multiple kernels simultaneously. This way, we automatically search over families of possible amino acid motifs. We compare our automated approach to three other predictors on four different datasets, and show that we perform better than the current state of the art. Further, our method provides some insights as to which sequence motifs are most useful for determining subcellular ocalization, which are in agreement with biological reasoning.

ei

PDF PDF Web DOI [BibTex]

PDF PDF Web DOI [BibTex]


no image
Discriminative K-means for Clustering

Ye, J., Zhao, Z., Wu, M.

In Advances in neural information processing systems 20, pages: 1649-1656, (Editors: Platt, J. C., D. Koller, Y. Singer, S. Roweis), Curran, Red Hook, NY, USA, Twenty-First Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
We present a theoretical study on the discriminative clustering framework, recently proposed for simultaneous subspace selection via linear discriminant analysis (LDA) and clustering. Empirical results have shown its favorable performance in comparison with several other popular clustering algorithms. However, the inherent relationship between subspace selection and clustering in this framework is not well understood, due to the iterative nature of the algorithm. We show in this paper that this iterative subspace selection and clustering is equivalent to kernel K-means with a specific kernel Gram matrix. This provides significant and new insights into the nature of this subspace selection procedure. Based on this equivalence relationship, we propose the Discriminative K-means (DisKmeans) algorithm for simultaneous LDA subspace selection and clustering, as well as an automatic parameter estimation procedure. We also present the nonlinear extension of DisKmeans using kernels. We show that the learning of the ke rnel matrix over a convex set of pre-specified kernel matrices can be incorporated into the clustering formulation. The connection between DisKmeans and several other clustering algorithms is also analyzed. The presented theories and algorithms are evaluated through experiments on a collection of benchmark data sets.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Bayesian Inference for Spiking Neuron Models with a Sparsity Prior

Gerwinn, S., Macke, J., Seeger, M., Bethge, M.

In Advances in neural information processing systems 20, pages: 529-536, (Editors: Platt, J. C., D. Koller, Y. Singer, S. Roweis), Curran, Red Hook, NY, USA, Twenty-First Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
Generalized linear models are the most commonly used tools to describe the stimulus selectivity of sensory neurons. Here we present a Bayesian treatment of such models. Using the expectation propagation algorithm, we are able to approximate the full posterior distribution over all weights. In addition, we use a Laplacian prior to favor sparse solutions. Therefore, stimulus features that do not critically influence neural activity will be assigned zero weights and thus be effectively excluded by the model. This feature selection mechanism facilitates both the interpretation of the neuron model as well as its predictive abilities. The posterior distribution can be used to obtain confidence intervals which makes it possible to assess the statistical significance of the solution. In neural data analysis, the available amount of experimental measurements is often limited whereas the parameter space is large. In such a situation, both regularization by a sparsity prior and uncertainty estimates for the model parameters are essential. We apply our method to multi-electrode recordings of retinal ganglion cells and use our uncertainty estimate to test the statistical significance of functional couplings between neurons. Furthermore we used the sparsity of the Laplace prior to select those filters from a spike-triggered covariance analysis that are most informative about the neural response.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Policy Gradients with Parameter-based Exploration for Control

Sehnke, F., Osendorfer, C., Rückstiess, T., Graves, A., Peters, J., Schmidhuber, J.

In ICANN 2008, pages: 387-396, (Editors: Kurkova-Pohlova, V. , R. Neruda, J. Koutnik), Springer, Berlin, Germany, 18th International Conference on Artificial Neural Networks, September 2008 (inproceedings)

Abstract
We present a model-free reinforcement learning method for partially observable Markov decision problems. Our method estimates a likelihood gradient by sampling directly in parameter space, which leads to lower variance gradient estimates than those obtained by policy gradient methods such as REINFORCE. For several complex control tasks, including robust standing with a humanoid robot, we show that our method outperforms well-known algorithms from the fields of policy gradients, finite difference methods and population based heuristics. We also provide a detailed analysis of the differences between our method and the other algorithms.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Distribution-free Learning of Bayesian Network Structure

Sun, X.

In ECML PKDD 2008, pages: 423-439, (Editors: Daelemans, W. , B. Goethals, K. Morik), Springer, Berlin, Germany, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, September 2008 (inproceedings)

Abstract
We present an independence-based method for learning Bayesian network (BN) structure without making any assumptions on the probability distribution of the domain. This is mainly useful for continuous domains. Even mixed continuous-categorical domains and structures containing vectorial variables can be handled. We address the problem by developing a non-parametric conditional independence test based on the so-called kernel dependence measure, which can be readily used by any existing independence-based BN structure learning algorithm. We demonstrate the structure learning of graphical models in continuous and mixed domains from real-world data without distributional assumptions. We also experimentally show that our test is a good alternative, in particular in case of small sample sizes, compared to existing tests, which can only be used in purely categorical or continuous domains.

ei

PDF PDF DOI [BibTex]

PDF PDF DOI [BibTex]


no image
Consistent Minimization of Clustering Objective Functions

von Luxburg, U., Bubeck, S., Jegelka, S., Kaufmann, M.

In Advances in neural information processing systems 20, pages: 961-968, (Editors: Platt, J. C., D. Koller, Y. Singer, S. Roweis), Curran, Red Hook, NY, USA, Twenty-First Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
Clustering is often formulated as a discrete optimization problem. The objective is to find, among all partitions of the data set, the best one according to some quality measure. However, in the statistical setting where we assume that the finite data set has been sampled from some underlying space, the goal is not to find the best partition of the given sample, but to approximate the true partition of the underlying space. We argue that the discrete optimization approach usually does not achieve this goal. As an alternative, we suggest the paradigm of nearest neighbor clustering‘‘. Instead of selecting the best out of all partitions of the sample, it only considers partitions in some restricted function class. Using tools from statistical learning theory we prove that nearest neighbor clustering is statistically consistent. Moreover, its worst case complexity is polynomial by co nstructi on, and it can b e implem ented wi th small average case co mplexity using b ranch an d bound.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
A Kernel Statistical Test of Independence

Gretton, A., Fukumizu, K., Teo, C., Song, L., Schölkopf, B., Smola, A.

In Advances in neural information processing systems 20, pages: 585-592, (Editors: JC Platt and D Koller and Y Singer and S Roweis), Curran, Red Hook, NY, USA, 21st Annual Conference on Neural Information Processing Systems (NIPS), September 2008 (inproceedings)

Abstract
Whereas kernel measures of independence have been widely applied in machine learning (notably in kernel ICA), there is as yet no method to determine whether they have detected statistically significant dependence. We provide a novel test of the independence hypothesis for one particular kernel independence measure, the Hilbert-Schmidt independence criterion (HSIC). The resulting test costs O(m^2), where m is the sample size. We demonstrate that this test outperforms established contingency table-based tests. Finally, we show the HSIC test also applies to text (and to structured data more generally), for which no other independence test presently exists.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Fitness Expectation Maximization

Wierstra, D., Schaul, T., Peters, J., Schmidhuber, J.

In PPSN 2008, pages: 337-346, (Editors: Rudolph, G. , T. Jansen, S. Lucas, C. Poloni, N. Beume), Springer, Berlin, Germany, 10th International Conference on Parallel Problem Solving From Nature, September 2008 (inproceedings)

Abstract
We present Fitness Expectation Maximization (FEM), a novel method for performing ‘black box’ function optimization. FEM searches the fitness landscape of an objective function using an instantiation of the well-known Expectation Maximization algorithm, producing search points to match the sample distribution weighted according to higher expected fitness. FEM updates both candidate solution parameters and the search policy, which is represented as a multinormal distribution. Inheriting EM’s stability and strong guarantees, the method is both elegant and competitive with some of the best heuristic search methods in the field, and performs well on a number of unimodal and multimodal benchmark tasks. To illustrate the potential practical applications of the approach, we also show experiments on finding the parameters for a controller of the challenging non-Markovian double pole balancing task.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Partial Least Squares Regression for Graph Mining

Saigo, H., Krämer, N., Tsuda, K.

In KDD2008, pages: 578-586, (Editors: Li, Y. , B. Liu, S. Sarawagi), ACM Press, New York, NY, USA, 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, August 2008 (inproceedings)

Abstract
Attributed graphs are increasingly more common in many application domains such as chemistry, biology and text processing. A central issue in graph mining is how to collect informative subgraph patterns for a given learning task. We propose an iterative mining method based on partial least squares regression (PLS). To apply PLS to graph data, a sparse version of PLS is developed first and then it is combined with a weighted pattern mining algorithm. The mining algorithm is iteratively called with different weight vectors, creating one latent component per one mining call. Our method, graph PLS, is efficient and easy to implement, because the weight vector is updated with elementary matrix calculations. In experiments, our graph PLS algorithm showed competitive prediction accuracies in many chemical datasets and its efficiency was significantly superior to graph boosting (gboost) and the naive method based on frequent graph mining.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Semi-Supervised Laplacian Regularization of Kernel Canonical Correlation Analysis

Blaschko, M., Lampert, C., Gretton, A.

In ECML PKDD 2008, pages: 133-145, (Editors: Daelemans, W. , B. Goethals, K. Morik), Springer, Berlin, Germany, 19th European Conference on Machine Learning, August 2008 (inproceedings)

Abstract
Kernel canonical correlation analysis (KCCA) is a dimensionality reduction technique for paired data. By finding directions that maximize correlation, KCCA learns representations that are more closely tied to the underlying semantics of the data rather than noise. However, meaningful directions are not only those that have high correlation to another modality, but also those that capture the manifold structure of the data. We propose a method that is simultaneously able to find highly correlated directions that are also located on high variance directions along the data manifold. This is achieved by the use of semi-supervised Laplacian regularization of KCCA. We show experimentally that Laplacian regularized training improves class separation over KCCA with only Tikhonov regularization, while causing no degradation in the correlation between modalities. We propose a model selection criterion based on the Hilbert-Schmidt norm of the semi-supervised Laplacian regularized cross-covariance operator, which we compute in closed form.

ei

PDF PDF DOI [BibTex]

PDF PDF DOI [BibTex]


no image
RKHS Representation of Measures Applied to Homogeneity, Independence, and Fourier Optics

Schölkopf, B., Sriperumbudur, B., Gretton, A., Fukumizu, K.

In OWR 2008, pages: 42-44, (Editors: K Jetter and S Smale and D-X Zhou), Mathematisches Forschungsinstitut, Oberwolfach-Walke, Germany, 30. Oberwolfach Report, August 2008 (inproceedings)

ei

PDF PDF [BibTex]

PDF PDF [BibTex]


no image
Learning Robot Dynamics for Computed Torque Control Using Local Gaussian Processes Regression

Nguyen-Tuong, D., Peters, J.

In LAB-RS 2008, pages: 59-64, (Editors: Stoica, A. , E. Tunstel, T. Huntsberger, T. Arslan, S. Vijayakumar, A. O. El-Rayis), IEEE Computer Society, Los Alamitos, CA, USA, 2008 ECSIS Symposium on Learning and Adaptive Behaviors for Robotic Systems, August 2008 (inproceedings)

Abstract
Accurate models of the robot dynamics allow the design of significantly more precise, energy-efficient and more compliant computed torque control for robots. However, in some cases the accuracy of rigid-body models does not suffice for sound control performance due to unmodeled nonlinearities such as hydraulic cables, complex friction, or actuator dynamics. In such cases, learning the models from data poses an interesting alternative and estimating the dynamics model using regression techniques becomes an important problem. However, the most accurate regression methods, e.g. Gaussian processes regression (GPR) and support vector regression (SVR), suffer from exceptional high computational complexity which prevents their usage for large numbers of samples or online learning to date. We proposed an approximation to the standard GPR using local Gaussian processes models. Due to reduced computational cost, local Gaussian processes (LGP) is capable for an online learning. Comparisons with other nonparametric regre ssions, e.g. standard GPR, SVR and locally weighted projection regression (LWPR), show that LGP has higher accuracy than LWPR and close to the performance of standard GPR and SVR while being sufficiently fast for online learning.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
A Decoupled Approach to Exemplar-based Unsupervised Learning

Nowozin, S., BakIr, G.

In ICML 2008, pages: 704-711, (Editors: Cohen, W. W., A. McCallum, S. Roweis), ACM Press, New York, NY, USA, 25th International Conference on Machine Learning, July 2008 (inproceedings)

Abstract
A recent trend in exemplar based unsupervised learning is to formulate the learning problem as a convex optimization problem. Convexity is achieved by restricting the set of possible prototypes to training exemplars. In particular, this has been done for clustering, vector quantization and mixture model density estimation. In this paper we propose a novel algorithm that is theoretically and practically superior to these convex formulations. This is possible by posing the unsupervised learning problem as a single convex master problem" with non-convex subproblems. We show that for the above learning tasks the subproblems are extremely wellbehaved and can be solved efficiently.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Relating clustering stability to properties of cluster boundaries

Ben-David, S., von Luxburg, U.

In COLT 2008, pages: 379-390, (Editors: Servedio, R. A., T. Zhang), Omnipress, Madison, WI, USA, 21st Annual Conference on Learning Theory, July 2008 (inproceedings)

Abstract
In this paper, we investigate stability-based methods for cluster model selection, in particular to select the number K of clusters. The scenario under consideration is that clustering is performed by minimizing a certain clustering quality function, and that a unique global minimizer exists. On the one hand we show that stability can be upper bounded by certain properties of the optimal clustering, namely by the mass in a small tube around the cluster boundaries. On the other hand, we provide counterexamples which show that a reverse statement is not true in general. Finally, we give some examples and arguments why, from a theoretic point of view, using clustering stability in a high sample setting can be problematic. It can be seen that distribution-free guarantees bounding the difference between the finite sample stability and the “true stability” cannot exist, unless one makes strong assumptions on the underlying distribution.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Compressed Sensing and Bayesian Experimental Design

Seeger, M., Nickisch, H.

In ICML 2008, pages: 912-919, (Editors: Cohen, W. W., A. McCallum, S. Roweis), ACM Press, New York, NY, USA, 25th International Conference on Machine Learning, July 2008 (inproceedings)

Abstract
We relate compressed sensing (CS) with Bayesian experimental design and provide a novel efficient approximate method for the latter, based on expectation propagation. In a large comparative study about linearly measuring natural images, we show that the simple standard heuristic of measuring wavelet coefficients top-down systematically outperforms CS methods using random measurements; the sequential projection optimisation approach of (Ji & Carin, 2007) performs even worse. We also show that our own approximate Bayesian method is able to learn measurement filters on full images efficiently which ouperform the wavelet heuristic. To our knowledge, ours is the first successful attempt at "learning compressed sensing" for images of realistic size. In contrast to common CS methods, our framework is not restricted to sparse signals, but can readily be applied to other notions of signal complexity or noise models. We give concrete ideas how our method can be scaled up to large signal representations.

ei

PDF PDF Web DOI [BibTex]

PDF PDF Web DOI [BibTex]


no image
Tailoring density estimation via reproducing kernel moment matching

Song, L., Zhang, X., Smola, A., Gretton, A., Schölkopf, B.

In Proceedings of the 25th International Conference onMachine Learning, pages: 992-999, (Editors: WW Cohen and A McCallum and S Roweis), ACM Press, New York, NY, USA, ICML, July 2008 (inproceedings)

Abstract
Moment matching is a popular means of parametric density estimation. We extend this technique to nonparametric estimation of mixture models. Our approach works by embedding distributions into a reproducing kernel Hilbert space, and performing moment matching in that space. This allows us to tailor density estimators to a function class of interest (i.e., for which we would like to compute expectations). We show our density estimation approach is useful in applications such as message compression in graphical models, and image classification and retrieval.

ei

PDF PDF DOI [BibTex]

PDF PDF DOI [BibTex]


no image
Injective Hilbert Space Embeddings of Probability Measures

Sriperumbudur, B., Gretton, A., Fukumizu, K., Lanckriet, G., Schölkopf, B.

In Proceedings of the 21st Annual Conference on Learning Theory, pages: 111-122, (Editors: RA Servedio and T Zhang), Omnipress, Madison, WI, USA, 21st Annual Conference on Learning Theory (COLT), July 2008 (inproceedings)

Abstract
A Hilbert space embedding for probability measures has recently been proposed, with applications including dimensionality reduction, homogeneity testing and independence testing. This embedding represents any probability measure as a mean element in a reproducing kernel Hilbert space (RKHS). The embedding function has been proven to be injective when the reproducing kernel is universal. In this case, the embedding induces a metric on the space of probability distributions defined on compact metric spaces. In the present work, we consider more broadly the problem of specifying characteristic kernels, defined as kernels for which the RKHS embedding of probability measures is injective. In particular, characteristic kernels can include non-universal kernels. We restrict ourselves to translation-invariant kernels on Euclidean space, and define the associated metric on probability measures in terms of the Fourier spectrum of the kernel and characteristic functions of these measures. The support of the kernel spectrum is important in finding whether a kernel is characteristic: in particular, the embedding is injective if and only if the kernel spectrum has the entire domain as its support. Characteristic kernels may nonetheless have difficulty in distinguishing certain distributions on the basis of finite samples, again due to the interaction of the kernel spectrum and the characteristic functions of the measures.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
A Hilbert-Schmidt Dependence Maximization Approach to Unsupervised Structure Discovery

Blaschko, M., Gretton, A.

In MLG 2008, pages: 1-3, 6th International Workshop on Mining and Learning with Graphs, July 2008 (inproceedings)

Abstract
In recent work by (Song et al., 2007), it has been proposed to perform clustering by maximizing a Hilbert-Schmidt independence criterion with respect to a predefined cluster structure Y , by solving for the partition matrix, II. We extend this approach here to the case where the cluster structure Y is not fixed, but is a quantity to be optimized; and we use an independence criterion which has been shown to be more sensitive at small sample sizes (the Hilbert-Schmidt Normalized Information Criterion, or HSNIC, Fukumizu et al., 2008). We demonstrate the use of this framework in two scenarios. In the first, we adopt a cluster structure selection approach in which the HSNIC is used to select a structure from several candidates. In the second, we consider the case where we discover structure by directly optimizing Y.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Adaptive Importance Sampling with Automatic Model Selection in Value Function Approximation

Hachiya, H., Akiyama, T., Sugiyama, M., Peters, J.

In AAAI 2008, pages: 1351-1356, (Editors: Fox, D. , C. P. Gomes), AAAI Press, Menlo Park, CA, USA, Twenty-Third Conference on Artificial Intelligence, July 2008 (inproceedings)

Abstract
Off-policy reinforcement learning is aimed at efficiently reusing data samples gathered in the past, which is an essential problem for physically grounded AI as experiments are usually prohibitively expensive. A common approach is to use importance sampling techniques for compensating for the bias caused by the difference between data-sampling policies and the target policy. However, existing off-policy methods do not often take the variance of value function estimators explicitly into account and therefore their performance tends to be unstable. To cope with this problem, we propose using an adaptive importance sampling technique which allows us to actively control the trade-off between bias and variance. We further provide a method for optimally determining the trade-off parameter based on a variant of cross-validation. We demonstrate the usefulness of the proposed approach through simulations.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Sparse Multiscale Gaussian Process Regression

Walder, C., Kim, K., Schölkopf, B.

In Proceedings of the 25th International Conference on Machine Learning, pages: 1112-1119, (Editors: WW Cohen and A McCallum and S Roweis), ACM Press, New York, NY, USA, ICML, July 2008 (inproceedings)

Abstract
Most existing sparse Gaussian process (g.p.) models seek computational advantages by basing their computations on a set of m basis functions that are the covariance function of the g.p. with one of its two inputs fixed. We generalise this for the case of Gaussian covariance function, by basing our computations on m Gaussian basis functions with arbitrary diagonal covariance matrices (or length scales). For a fixed number of basis functions and any given criteria, this additional flexibility permits approximations no worse and typically better than was previously possible. We perform gradient based optimisation of the marginal likelihood, which costs O(m2n) time where n is the number of data points, and compare the method to various other sparse g.p. methods. Although we focus on g.p. regression, the central idea is applicable to all kernel based algorithms, and we also provide some results for the support vector machine (s.v.m.) and kernel ridge regression (k.r.r.). Our approach outperforms the other methods, particularly for the case of very few basis functions, i.e. a very high sparsity ratio.

ei

PDF PDF DOI [BibTex]

PDF PDF DOI [BibTex]


no image
Example-Based Learning for Single-Image Super-Resolution

Kim, K., Kwon, Y.

In DAGM 2008, pages: 456-463, (Editors: Rigoll, G. ), Springer, Berlin, Germany, 30th Annual Symposium of the German Association for Pattern Recognition, June 2008 (inproceedings)

Abstract
This paper proposes a regression-based method for single-image super-resolution. Kernel ridge regression (KRR) is used to estimate the high-frequency details of the underlying high-resolution image. A sparse solution of KRR is found by combining the ideas of kernel matching pursuit and gradient descent, which allows time-complexity to be kept to a moderate level. To resolve the problem of ringing artifacts occurring due to the regularization effect, the regression results are post-processed using a prior model of a generic image class. Experimental results demonstrate the effectiveness of the proposed method.

ei

PDF DOI [BibTex]

PDF DOI [BibTex]


no image
A Multiple Kernel Learning Approach to Joint Multi-Class Object Detection

Lampert, C., Blaschko, M.

In DAGM 2008, pages: 31-40, (Editors: Rigoll, G. ), Springer, Berlin, Germany, 30th Annual Symposium of the German Association for Pattern Recognition, June 2008, Main Award DAGM 2008 (inproceedings)

Abstract
Most current methods for multi-class object classification and localization work as independent 1-vs-rest classifiers. They decide whether and where an object is visible in an image purely on a per-class basis. Joint learning of more than one object class would generally be preferable, since this would allow the use of contextual information such as co-occurrence between classes. However, this approach is usually not employed because of its computational cost. In this paper we propose a method to combine the efficiency of single class localization with a subsequent decision process that works jointly for all given object classes. By following a multiple kernel learning (MKL) approach, we automatically obtain a sparse dependency graph of relevant object classes on which to base the decision. Experiments on the PASCAL VOC 2006 and 2007 datasets show that the subsequent joint decision step clearly improves the accuracy compared to single class detection.

ei

PDF ZIP Web DOI [BibTex]

PDF ZIP Web DOI [BibTex]


no image
Natural Evolution Strategies

Wierstra, D., Schaul, T., Peters, J., Schmidhuber, J.

In CEC 2008, pages: 3381-3387, IEEE, Piscataway, NJ, USA, IEEE Congress on Evolutionary Computation, June 2008 (inproceedings)

Abstract
This paper presents natural evolution strategies (NES), a novel algorithm for performing real-valued dasiablack boxpsila function optimization: optimizing an unknown objective function where algorithm-selected function measurements constitute the only information accessible to the method. Natural evolution strategies search the fitness landscape using a multivariate normal distribution with a self-adapting mutation matrix to generate correlated mutations in promising regions. NES shares this property with covariance matrix adaption (CMA), an evolution strategy (ES) which has been shown to perform well on a variety of high-precision optimization tasks. The natural evolution strategies algorithm, however, is simpler, less ad-hoc and more principled. Self-adaptation of the mutation matrix is derived using a Monte Carlo estimate of the natural gradient towards better expected fitness. By following the natural gradient instead of the dasiavanillapsila gradient, we can ensure efficient update steps while preventing early convergence due to overly greedy updates, resulting in reduced sensitivity to local suboptima. We show NES has competitive performance with CMA on unimodal tasks, while outperforming it on several multimodal tasks that are rich in deceptive local optima.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Partitioning of Image Datasets using Discriminative Context Information

Lampert, CH.

In CVPR 2008, pages: 1-8, IEEE Computer Society, Los Alamitos, CA, USA, IEEE Computer Society Conference on Computer Vision and Pattern Recognition, June 2008 (inproceedings)

Abstract
We propose a new method to partition an unlabeled dataset, called Discriminative Context Partitioning (DCP). It is motivated by the idea of splitting the dataset based only on how well the resulting parts can be separated from a context class of disjoint data points. This is in contrast to typical clustering techniques like K-means that are based on a generative model by implicitly or explicitly searching for modes in the distribution of samples. The discriminative criterion in DCP avoids the problems that density based methods have when the a priori assumption of multimodality is violated, when the number of samples becomes small in relation to the dimensionality of the feature space, or if the cluster sizes are strongly unbalanced. We formulate DCP‘s separation property as a large-margin criterion, and show how the resulting optimization problem can be solved efficiently. Experiments on the MNIST and USPS datasets of handwritten digits and on a subset of the Caltech256 dataset show that, given a suitable context, DCP can achieve good results even in situation where density-based clustering techniques fail.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Correlational Spectral Clustering

Blaschko, MB., Lampert, CH.

In CVPR 2008, pages: 1-8, IEEE Computer Society, Los Alamitos, CA, USA, IEEE Computer Society Conference on Computer Vision and Pattern Recognition, June 2008 (inproceedings)

Abstract
We present a new method for spectral clustering with paired data based on kernel canonical correlation analysis, called correlational spectral clustering. Paired data are common in real world data sources, such as images with text captions. Traditional spectral clustering algorithms either assume that data can be represented by a single similarity measure, or by co-occurrence matrices that are then used in biclustering. In contrast, the proposed method uses separate similarity measures for each data representation, and allows for projection of previously unseen data that are only observed in one representation (e.g. images but not text). We show that this algorithm generalizes traditional spectral clustering algorithms and show consistent empirical improvement over spectral clustering on a variety of datasets of images with associated text.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Approximate Dynamic Programming with Gaussian Processes

Deisenroth, M., Peters, J., Rasmussen, C.

In ACC 2008, pages: 4480-4485, IEEE Service Center, Piscataway, NJ, USA, 2008 American Control Conference, June 2008 (inproceedings)

Abstract
In general, it is difficult to determine an optimal closed-loop policy in nonlinear control problems with continuous-valued state and control domains. Hence, approximations are often inevitable. The standard method of discretizing states and controls suffers from the curse of dimensionality and strongly depends on the chosen temporal sampling rate. In this paper, we introduce Gaussian process dynamic programming (GPDP) and determine an approximate globally optimal closed-loop policy. In GPDP, value functions in the Bellman recursion of the dynamic programming algorithm are modeled using Gaussian processes. GPDP returns an optimal statefeedback for a finite set of states. Based on these outcomes, we learn a possibly discontinuous closed-loop policy on the entire state space by switching between two independently trained Gaussian processes. A binary classifier selects one Gaussian process to predict the optimal control signal. We show that GPDP is able to yield an almost optimal solution to an LQ problem using few sample points. Moreover, we successfully apply GPDP to the underpowered pendulum swing up, a complex nonlinear control problem.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Beyond Sliding Windows: Object Localization by Efficient Subwindow Search

Lampert, C., Blaschko, M., Hofmann, T.

In CVPR 2008, pages: 1-8, IEEE Computer Society, Los Alamitos, CA, USA, IEEE Computer Society Conference on Computer Vision and Pattern Recognition, June 2008, Best paper award (inproceedings)

Abstract
Most successful object recognition systems rely on binary classification, deciding only if an object is present or not, but not providing information on the actual object location. To perform localization, one can take a sliding window approach, but this strongly increases the computational cost, because the classifier function has to be evaluated over a large set of candidate subwindows. In this paper, we propose a simple yet powerful branchand- bound scheme that allows efficient maximization of a large class of classifier functions over all possible subimages. It converges to a globally optimal solution typically in sublinear time. We show how our method is applicable to different object detection and retrieval scenarios. The achieved speedup allows the use of classifiers for localization that formerly were considered too slow for this task, such as SVMs with a spatial pyramid kernel or nearest neighbor classifiers based on the 2-distance. We demonstrate state-of-the-art performance of the resulting systems on the UIUC Cars dataset, the PASCAL VOC 2006 dataset and in the PASCAL VOC 2007 competition.

ei

PDF PDF Web DOI [BibTex]

PDF PDF Web DOI [BibTex]


no image
Computed Torque Control with Nonparametric Regression Models

Nguyen-Tuong, D., Seeger, M., Peters, J.

In ACC 2008, pages: 212-217, IEEE Service Center, Piscataway, NJ, USA, 2008 American Control Conference, June 2008 (inproceedings)

Abstract
Computed torque control allows the design of considerably more precise, energy-efficient and compliant controls for robots. However, the major obstacle is the requirement of an accurate model for torque generation, which cannot be obtained in some cases using rigid-body formulations due to unmodeled nonlinearities, such as complex friction or actuator dynamics. In such cases, models approximated from robot data present an appealing alternative. In this paper, we compare two nonparametric regression methods for model approximation, i.e., locally weighted projection regression (LWPR) and Gaussian process regression (GPR). While locally weighted regression was employed for real-time model estimation in learning adaptive control, Gaussian process regression has not been used in control to-date due to high computational requirements. The comparison includes the assessment of model approximation for both regression methods using data originated from SARCOS robot arm, as well as an evaluation of the robot tracking p erformance in computed torque control employing the approximated models. Our results show that GPR can be applied for real-time control achieving higher accuracy. However, for the online learning LWPR is superior by reason of lower computational requirements.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
Multi-Classification by Categorical Features via Clustering

Seldin, Y., Tishby, N.

In In the proceedings of the 25th International Conference on Machine Learning (ICML 2008), pages: 920-927, 25th International Conference on Machine Learning (ICML), June 2008 (inproceedings)

Abstract
We derive a generalization bound for multi-classification schemes based on grid clustering in categorical parameter product spaces. Grid clustering partitions the parameter space in the form of a Cartesian product of partitions for each of the parameters. The derived bound provides a means to evaluate clustering solutions in terms of the generalization power of a built-on classifier. For classification based on a single feature the bound serves to find a globally optimal classification rule. Comparison of the generalization power of individual features can then be used for feature ranking. Our experiments show that in this role the bound is much more precise than mutual information or normalized correlation indices.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
A Kernel Test of Nonlinear Granger Causality

Sun, X.

In Proceedings of the Workshop on Inference and Estimation in Probabilistic Time-Series Models, pages: 79-89, (Editors: Barber, D. , A. T. Cemgil, S. Chiappa), Isaac Newton Institute for Mathematical Sciences, Cambridge, United Kingdom, Workshop on Inference and Estimation in Probabilistic Time-Series Models, June 2008 (inproceedings)

Abstract
We present a novel test of nonlinear Granger causality in bivariate time series. The trace norm of conditional covariance operators is used to capture the prediction errors. Based on this measure, a subsampling-based multiple testing procedure tests the prediction improvement of one time series by the other one. The distributional properties of the resulting p-values reveal the direction of Granger causality. Encouraging results of experiments with simulated and real-world data support our approach.

ei

PDF [BibTex]

PDF [BibTex]


Thumb xl teaser
Bayesian Color Constancy Revisited

Gehler, P., Rother, C., Blake, A., Minka, T., Sharp, T.

In IEEE Computer Society Conference on Computer Vision and Pattern Recognition, CVPR, June 2008, http://dx.doi.org/10.1109/CVPR.2008.4587765 (inproceedings)

ei

website+code+data pdf [BibTex]

website+code+data pdf [BibTex]


no image
Real-time Learning of Resolved Velocity Control on a Mitsubishi PA-10

Peters, J., Nguyen-Tuong, D.

In ICRA 2008, pages: 2872-2877, IEEE Service Center, Piscataway, NJ, USA, 2008 IEEE International Conference on Robotics and Automation, May 2008 (inproceedings)

Abstract
Learning inverse kinematics has long been fascinating the robot learning community. While humans acquire this transformation to complicated tool spaces with ease, it is not a straightforward application for supervised learning algorithms due to non-convex learning problem. However, the key insight that the problem can be considered convex in small local regions allows the application of locally linear learning methods. Nevertheless, the local solution of the problem depends on the data distribution which can result into inconsistent global solutions with large model discontinuities. While this problem can be treated in various ways in offline learning, it poses a serious problem for online learning. Previous approaches to the real-time learning of inverse kinematics avoid this problem using smart data generation, such as the learner biasses its own solution. Such biassed solutions can result into premature convergence, and from the resulting solution it is often hard to understand what has been learned in tha t local region. This paper improves and solves this problem by presenting a learning algorithm which can deal with this inconsistency through re-weighting the data online. Furthermore, we show that our algorithms work not only in simulation, but we present real-time learning results on a physical Mitsubishi PA-10 robot arm.

ei

PDF Web DOI [BibTex]

PDF Web DOI [BibTex]


no image
New Frontiers in Characterizing Structure and Dynamics by NMR

Nilges, M., Markwick, P., Malliavin, TE., Rieping, W., Habeck, M.

In Computational Structural Biology: Methods and Applications, pages: 655-680, (Editors: Schwede, T. , M. C. Peitsch), World Scientific, New Jersey, NJ, USA, May 2008 (inbook)

Abstract
Nuclear Magnetic Resonance (NMR) spectroscopy has emerged as the method of choice for studying both the structure and the dynamics of biological macromolecule in solution. Despite the maturity of the NMR method for structure determination, its application faces a number of challenges. The method is limited to systems of relatively small molecular mass, data collection times are long, data analysis remains a lengthy procedure, and it is difficult to evaluate the quality of the final structures. The last years have seen significant advances in experimental techniques to overcome or reduce some limitations. The function of bio-macromolecules is determined by both their 3D structure and conformational dynamics. These molecules are inherently flexible systems displaying a broad range of dynamics on time–scales from picoseconds to seconds. NMR is unique in its ability to obtain dynamic information on an atomic scale. The experimental information on structure and dynamics is intricately mixed. It is however difficult to unite both structural and dynamical information into one consistent model, and protocols for the determination of structure and dynamics are performed independently. This chapter deals with the challenges posed by the interpretation of NMR data on structure and dynamics. We will first relate the standard structure calculation methods to Bayesian probability theory. We will then briefly describe the advantages of a fully Bayesian treatment of structure calculation. Then, we will illustrate the advantages of using Bayesian reasoning at least partly in standard structure calculations. The final part will be devoted to interpretation of experimental data on dynamics.

ei

Web [BibTex]

Web [BibTex]


no image
Graph Mining with Variational Dirichlet Process Mixture Models

Tsuda, K., Kurihara, K.

In SDM 2008, pages: 432-442, (Editors: Zaki, M. J.), Society for Industrial and Applied Mathematics, Philadelphia, PA, USA, 8th SIAM International Conference on Data Mining, April 2008 (inproceedings)

Abstract
Graph data such as chemical compounds and XML documents are getting more common in many application domains. A main difficulty of graph data processing lies in the intrinsic high dimensionality of graphs, namely, when a graph is represented as a binary feature vector of indicators of all possible subgraph patterns, the dimensionality gets too large for usual statistical methods. We propose a nonparametric Bayesian method for clustering graphs and selecting salient patterns at the same time. Variational inference is adopted here, because sampling is not applicable due to extremely high dimensionality. The feature set minimizing the free energy is efficiently collected with the DFS code tree, where the generation of useless subgraphs is suppressed by a tree pruning condition. In experiments, our method is compared with a simpler approach based on frequent subgraph mining, and graph kernels.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Model-Based Reinforcement Learning with Continuous States and Actions

Deisenroth, M., Rasmussen, C., Peters, J.

In ESANN 2008, pages: 19-24, (Editors: Verleysen, M. ), d-side, Evere, Belgium, European Symposium on Artificial Neural Networks, April 2008 (inproceedings)

Abstract
Finding an optimal policy in a reinforcement learning (RL) framework with continuous state and action spaces is challenging. Approximate solutions are often inevitable. GPDP is an approximate dynamic programming algorithm based on Gaussian process (GP) models for the value functions. In this paper, we extend GPDP to the case of unknown transition dynamics. After building a GP model for the transition dynamics, we apply GPDP to this model and determine a continuous-valued policy in the entire state space. We apply the resulting controller to the underpowered pendulum swing up. Moreover, we compare our results on this RL task to a nearly optimal discrete DP solution in a fully known environment.

ei

PDF Web [BibTex]

PDF Web [BibTex]


no image
Learning Inverse Dynamics: A Comparison

Nguyen-Tuong, D., Peters, J., Seeger, M., Schölkopf, B.

In Advances in Computational Intelligence and Learning: Proceedings of the European Symposium on Artificial Neural Networks, pages: 13-18, (Editors: M Verleysen), d-side, Evere, Belgium, 16th European Symposium on Artificial Neural Networks (ESANN), April 2008 (inproceedings)

Abstract
While it is well-known that model can enhance the control performance in terms of precision or energy efficiency, the practical application has often been limited by the complexities of manually obtaining sufficiently accurate models. In the past, learning has proven a viable alternative to using a combination of rigid-body dynamics and handcrafted approximations of nonlinearities. However, a major open question is what nonparametric learning method is suited best for learning dynamics? Traditionally, locally weighted projection regression (LWPR), has been the standard method as it is capable of online, real-time learning for very complex robots. However, while LWPR has had significant impact on learning in robotics, alternative nonparametric regression methods such as support vector regression (SVR) and Gaussian processes regression (GPR) offer interesting alternatives with fewer open parameters and potentially higher accuracy. In this paper, we evaluate these three alternatives for model learning. Our comparison consists out of the evaluation of learning quality for each regression method using original data from SARCOS robot arm, as well as the robot tracking performance employing learned models. The results show that GPR and SVR achieve a superior learning precision and can be applied for real-time control obtaining higher accuracy. However, for the online learning LWPR presents the better method due to its lower computational requirements.

ei

PDF Web [BibTex]

PDF Web [BibTex]