Header logo is ei


2013


no image
Studying large-scale brain networks: electrical stimulation and neural-event-triggered fMRI

Logothetis, N., Eschenko, O., Murayama, Y., Augath, M., Steudel, T., Evrard, H., Besserve, M., Oeltermann, A.

Twenty-Second Annual Computational Neuroscience Meeting (CNS*2013), July 2013, journal = {BMC Neuroscience}, year = {2013}, month = {7}, volume = {14}, number = {Supplement 1}, pages = {A1}, (talk)

Web [BibTex]

2013

Web [BibTex]


no image
Coupling between spiking activity and beta band spatio-temporal patterns in the macaque PFC

Safavi, S., Panagiotaropoulos, T., Kapoor, V., Logothetis, N., Besserve, M.

43rd Annual Meeting of the Society for Neuroscience (Neuroscience), 2013 (poster)

[BibTex]

[BibTex]


no image
Gaussian Process Vine Copulas for Multivariate Dependence

Lopez-Paz, D., Hernandez-Lobato, J., Ghahramani, Z.

International Conference on Machine Learning (ICML), 2013 (poster)

PDF [BibTex]

PDF [BibTex]


no image
Domain Generalization via Invariant Feature Representation

Muandet, K., Balduzzi, D., Schölkopf, B.

30th International Conference on Machine Learning (ICML2013), 2013 (poster)

PDF [BibTex]

PDF [BibTex]


no image
Analyzing locking of spikes to spatio-temporal patterns in the macaque prefrontal cortex

Safavi, S., Panagiotaropoulos, T., Kapoor, V., Logothetis, N., Besserve, M.

Bernstein Conference, 2013 (poster)

DOI [BibTex]

DOI [BibTex]


no image
One-class Support Measure Machines for Group Anomaly Detection

Muandet, K., Schölkopf, B.

29th Conference on Uncertainty in Artificial Intelligence (UAI), 2013 (poster)

PDF [BibTex]

PDF [BibTex]


no image
The Randomized Dependence Coefficient

Lopez-Paz, D., Hennig, P., Schölkopf, B.

Neural Information Processing Systems (NIPS), 2013 (poster)

PDF [BibTex]

PDF [BibTex]


no image
Characterization of different types of sharp-wave ripple signatures in the CA1 of the macaque hippocampus

Ramirez-Villegas, J., Logothetis, N., Besserve, M.

4th German Neurophysiology PhD Meeting Networks, 2013 (poster)

Web [BibTex]

Web [BibTex]


no image
Domain Generalization via Invariant Feature Representation

Muandet, K.

30th International Conference on Machine Learning (ICML2013), 2013 (talk)

PDF [BibTex]

PDF [BibTex]

2009


no image
Machine Learning for Brain-Computer Interfaces

Hill, NJ.

Mini-Symposia on Assistive Machine Learning for People with Disabilities at NIPS (AMD), December 2009 (talk)

Abstract
Brain-computer interfaces (BCI) aim to be the ultimate in assistive technology: decoding a user‘s intentions directly from brain signals without involving any muscles or peripheral nerves. Thus, some classes of BCI potentially offer hope for users with even the most extreme cases of paralysis, such as in late-stage Amyotrophic Lateral Sclerosis, where nothing else currently allows communication of any kind. Other lines in BCI research aim to restore lost motor function in as natural a way as possible, reconnecting and in some cases re-training motor-cortical areas to control prosthetic, or previously paretic, limbs. Research and development are progressing on both invasive and non-invasive fronts, although BCI has yet to make a breakthrough to widespread clinical application. The high-noise high-dimensional nature of brain-signals, particularly in non-invasive approaches and in patient populations, make robust decoding techniques a necessity. Generally, the approach has been to use relatively simple feature extraction techniques, such as template matching and band-power estimation, coupled to simple linear classifiers. This has led to a prevailing view among applied BCI researchers that (sophisticated) machine-learning is irrelevant since "it doesn‘t matter what classifier you use once you‘ve done your preprocessing right and extracted the right features." I shall show a few examples of how this runs counter to both the empirical reality and the spirit of what needs to be done to bring BCI into clinical application. Along the way I‘ll highlight some of the interesting problems that remain open for machine-learners.

PDF Web Web [BibTex]

2009

PDF Web Web [BibTex]


no image
Learning Probabilistic Models via Bayesian Inverse Planning

Boularias, A., Chaib-Draa, B.

NIPS Workshop on Probabilistic Approaches for Robotics and Control, December 2009 (poster)

PDF Web [BibTex]

PDF Web [BibTex]


no image
Bayesian Quadratic Reinforcement Learning

Hennig, P., Stern, D., Graepel, T.

NIPS Workshop on Probabilistic Approaches for Robotics and Control, December 2009 (poster)

PDF Web [BibTex]

PDF Web [BibTex]


no image
PAC-Bayesian Approach to Formulation of Clustering Objectives

Seldin, Y.

NIPS Workshop on "Clustering: Science or Art? Towards Principled Approaches", December 2009 (talk)

Abstract
Clustering is a widely used tool for exploratory data analysis. However, the theoretical understanding of clustering is very limited. We still do not have a well-founded answer to the seemingly simple question of "how many clusters are present in the data?", and furthermore a formal comparison of clusterings based on different optimization objectives is far beyond our abilities. The lack of good theoretical support gives rise to multiple heuristics that confuse the practitioners and stall development of the field. We suggest that the ill-posed nature of clustering problems is caused by the fact that clustering is often taken out of its subsequent application context. We argue that one does not cluster the data just for the sake of clustering it, but rather to facilitate the solution of some higher level task. By evaluation of the clustering‘s contribution to the solution of the higher level task it is possible to compare different clusterings, even those obtained by different optimization objectives. In the preceding work it was shown that such an approach can be applied to evaluation and design of co-clustering solutions. Here we suggest that this approach can be extended to other settings, where clustering is applied.

PDF Web Web [BibTex]

PDF Web Web [BibTex]


no image
Semi-supervised Kernel Canonical Correlation Analysis of Human Functional Magnetic Resonance Imaging Data

Shelton, JA.

Women in Machine Learning Workshop (WiML), December 2009 (talk)

Abstract
Kernel Canonical Correlation Analysis (KCCA) is a general technique for subspace learning that incorporates principal components analysis (PCA) and Fisher linear discriminant analysis (LDA) as special cases. By finding directions that maximize correlation, KCCA learns representations tied more closely to underlying process generating the the data and can ignore high-variance noise directions. However, for data where acquisition in a given modality is expensive or otherwise limited, KCCA may suffer from small sample effects. We propose to use semi-supervised Laplacian regularization to utilize data that are present in only one modality. This manifold learning approach is able to find highly correlated directions that also lie along the data manifold, resulting in a more robust estimate of correlated subspaces. Functional magnetic resonance imaging (fMRI) acquired data are naturally amenable to subspace techniques as data are well aligned and such data of the human brain are a particularly interesting candidate. In this study we implemented various supervised and semi-supervised versions of KCCA on human fMRI data, with regression to single and multivariate labels (corresponding to video content subjects viewed during the image acquisition). In each variate condition, Laplacian regularization improved performance whereas the semi-supervised variants of KCCA yielded the best performance. We additionally analyze the weights learned by the regression in order to infer brain regions that are important during different types of visual processing.

PDF Web [BibTex]


no image
Policy Transfer in Apprenticeship Learning

Boularias, A., Chaib-Draa, B.

NIPS Workshop on Transfer Learning for Structured Data (TLSD-09), December 2009 (poster)

PDF Web [BibTex]

PDF Web [BibTex]


no image
Event-Related Potentials in Brain-Computer Interfacing

Hill, NJ.

Invited lecture on the bachelor & masters course "Introduction to Brain-Computer Interfacing", October 2009 (talk)

Abstract
An introduction to event-related potentials with specific reference to their use in brain-computer interfacing applications and research.

PDF [BibTex]

PDF [BibTex]


no image
BCI2000 and Python

Hill, NJ.

Invited lecture at the 5th International BCI2000 Workshop, October 2009 (talk)

Abstract
A tutorial, with exercises, on how to integrate your own Python code with the BCI2000 software package.

PDF [BibTex]

PDF [BibTex]


no image
Implementing a Signal Processing Filter in BCI2000 Using C++

Hill, NJ., Mellinger, J.

Invited lecture at the 5th International BCI2000 Workshop, October 2009 (talk)

Abstract
This tutorial shows how the functionality of the BCI2000 software package can be extended with one‘s own code, using BCI2000‘s C++ API.

PDF [BibTex]

PDF [BibTex]


no image
Clinical PET/MRI-System and Its Applications with MRI Based Attenuation Correction

Kolb, A., Hofmann, M., Sossi, V., Wehrl, H., Sauter, A., Schmid, A., Schlemmer, H., Claussen, C., Pichler, B.

IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC 2009), 2009, pages: 1, October 2009 (poster)

Abstract
Clinical PET/MRI is an emerging new hybrid imaging modality. In addition to provide an unique possibility for multifunctional imaging with temporally and spatially matched data, it also provides anatomical information that can also be used for attenuation correction with no radiation exposure to the subjects. A plus of combined compared to sequential PET and MR imaging is the reduction of total scan time. Here we present our initial experience with a hybrid brain PET/MRI system. Due to the ethical approval patient scans could only be performed after a diagnostic PET/CT. We estimate that in approximately 50% of the cases PET/MRI was of superior diagnostic value compared to PET/CT and was able to provide additional information, such as DTI, spectroscopy and Time Of Flight (TOF) angiography. Here we present 3 patient cases in oncology, a retropharyngeal carcinoma in neurooncology, a relapsing meningioma and in neurology a pharyngeal carcinoma in addition to an infraction of the right hemisphere. For quantitative PET imaging attenuation correction is obligatory. In current PET/MRI setup we used our MRI based atlas method for calculating the mu-map for attenuation correction. MR-based attenuation correction accuracy was quantitatively compared to CT-based PET attenuation correction. Extensive studies to assess potential mutual interferences between PET and MR imaging modalities as well as NEMA measurements have been performed. The first patient studies as well as the phantom tests clearly demonstrated the overall good imaging performance of this first human PET/MRI system. Ongoing work concentrates on advanced normalization and reconstruction methods incorporating count-rate based algorithms.

Web [BibTex]

Web [BibTex]


no image
A flowering-time gene network model for association analysis in Arabidopsis thaliana

Klotzbücher, K., Kobayashi, Y., Shervashidze, N., Borgwardt, K., Weigel, D.

2009(39):95-96, German Conference on Bioinformatics (GCB '09), September 2009 (poster)

Abstract
In our project we want to determine a set of single nucleotide polymorphisms (SNPs), which have a major effect on the flowering time of Arabidopsis thaliana. Instead of performing a genome-wide association study on all SNPs in the genome of Arabidopsis thaliana, we examine the subset of SNPs from the flowering-time gene network model. We are interested in how the results of the association study vary when using only the ascertained subset of SNPs from the flowering network model, and when additionally using the information encoded by the structure of the network model. The network model is compiled from the literature by manual analysis and contains genes which have been found to affect the flowering time of Arabidopsis thaliana [Far+08; KW07]. The genes in this model are annotated with the SNPs that are located in these genes, or in near proximity to them. In a baseline comparison between the subset of SNPs from the graph and the set of all SNPs, we omit the structural information and calculate the correlation between the individual SNPs and the flowering time phenotype by use of statistical methods. Through this we can determine the subset of SNPs with the highest correlation to the flowering time. In order to further refine this subset, we include the additional information provided by the network structure by conducting a graph-based feature pre-selection. In the further course of this project we want to validate and examine the resulting set of SNPs and their corresponding genes with experimental methods.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Initial Data from a first PET/MRI-System and its Applications in Clinical Studies Using MRI Based Attenuation Correction

Kolb, A., Hofmann, M., Sossi, V., Wehrl, H., Sauter, A., Schmid, A., Judenhofer, M., Schlemmer, H., Claussen, C., Pichler, B.

2009 World Molecular Imaging Congress, 2009, pages: 1200, September 2009 (poster)

Web [BibTex]

Web [BibTex]


no image
A High-Speed Object Tracker from Off-the-Shelf Components

Lampert, C., Peters, J.

First IEEE Workshop on Computer Vision for Humanoid Robots in Real Environments at ICCV 2009, 1, pages: 1, September 2009 (poster)

Abstract
We introduce RTblob, an open-source real-time vision system for 3D object detection that achieves over 200 Hz tracking speed with only off-the-shelf hardware component. It allows fast and accurate tracking of colored objects in 3D without expensive and often custom-built hardware, instead making use of the PC graphics cards for the necessary image processing operations.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Estimating Critical Stimulus Features from Psychophysical Data: The Decision-Image Technique Applied to Human Faces

Macke, J., Wichmann, F.

Journal of Vision, 9(8):31, 9th Annual Meeting of the Vision Sciences Society (VSS), August 2009 (poster)

Abstract
One of the main challenges in the sensory sciences is to identify the stimulus features on which the sensory systems base their computations: they are a pre-requisite for computational models of perception. We describe a technique---decision-images--- for extracting critical stimulus features based on logistic regression. Rather than embedding the stimuli in noise, as is done in classification image analysis, we want to infer the important features directly from physically heterogeneous stimuli. A Decision-image not only defines the critical region-of-interest within a stimulus but is a quantitative template which defines a direction in stimulus space. Decision-images thus enable the development of predictive models, as well as the generation of optimized stimuli for subsequent psychophysical investigations. Here we describe our method and apply it to data from a human face discrimination experiment. We show that decision-images are able to predict human responses not only in terms of overall percent correct but are able to predict, for individual observers, the probabilities with which individual faces are (mis-) classified. We then test the predictions of the models using optimized stimuli. Finally, we discuss possible generalizations of the approach and its relationships with other models.

Web DOI [BibTex]

Web DOI [BibTex]


no image
Semi-supervised Analysis of Human fMRI Data

Shelton, JA., Blaschko, MB., Lampert, CH., Bartels, A.

Berlin Brain Computer Interface Workshop on Advances in Neurotechnology, 2009, pages: 1, July 2009 (poster)

Abstract
Kernel Canonical Correlation Analysis (KCCA) is a general technique for subspace learning that incorporates principal components analysis (PCA) and Fisher linear discriminant analysis (LDA) as special cases. By finding directions that maximize correlation, CCA learns representations tied more closely to underlying process generating the the data and can ignore high-variance noise directions. However, for data where acquisition in a given modality is expensive or otherwise limited, CCA may suffer from small sample effects. We propose to use semisupervised Laplacian regularization to utilize data that are present in only one modality. This approach is able to find highly correlated directions that also lie along the data manifold, resulting in a more robust estimate of correlated subspaces. Functional magnetic resonance imaging (fMRI) acquired data are naturally amenable to subspace techniques as data are well aligned. fMRI data of the human brain are a particularly interesting candidate. In this study we implemented various supervised and semi-supervised versions of CCA on human fMRI data, with regression to single and multivariate labels (corresponding to video content subjects viewed during the image acquisition). In each variate condition, the semi-supervised variants of CCA performed better than the supervised variants, including a supervised variant with Laplacian regularization. We additionally analyze the weights learned by the regression in order to infer brain regions that are important to different types of visual processing.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Randomized algorithms for statistical image analysis based on percolation theory

Davies, P., Langovoy, M., Wittich, O.

27th European Meeting of Statisticians (EMS), July 2009 (talk)

Abstract
We propose a novel probabilistic method for detection of signals and reconstruction of images in the presence of random noise. The method uses results from percolation and random graph theories (see Grimmett (1999)). We address the problem of detection and estimation of signals in situations where the signal-to-noise ratio is particularly low. We present an algorithm that allows to detect objects of various shapes in noisy images. The algorithm has linear complexity and exponential accuracy. Our algorithm substantially di ers from wavelets-based algorithms (see Arias-Castro et.al. (2005)). Moreover, we present an algorithm that produces a crude estimate of an object based on the noisy picture. This algorithm also has linear complexity and is appropriate for real-time systems. We prove results on consistency and algorithmic complexity of our procedures.

Web PDF [BibTex]

Web PDF [BibTex]


no image
Learning Motor Primitives for Robotics

Kober, J., Peters, J., Oztop, E.

Advanced Telecommunications Research Center ATR, June 2009 (talk)

Abstract
The acquisition and self-improvement of novel motor skills is among the most important problems in robotics. Motor primitives offer one of the most promising frameworks for the application of machine learning techniques in this context. Employing the Dynamic Systems Motor primitives originally introduced by Ijspeert et al. (2003), appropriate learning algorithms for a concerted approach of both imitation and reinforcement learning are presented. Using these algorithms new motor skills, i.e., Ball-in-a-Cup, Ball-Paddling and Dart-Throwing, are learned.

[BibTex]

[BibTex]


no image
Learning To Detect Unseen Object Classes by Between-Class Attribute Transfer

Lampert, C.

IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), June 2009 (talk)

Web [BibTex]

Web [BibTex]


no image
Optimization of k-Space Trajectories by Bayesian Experimental Design

Seeger, M., Nickisch, H., Pohmann, R., Schölkopf, B.

17(2627), 17th Annual Meeting of the International Society for Magnetic Resonance in Medicine (ISMRM), April 2009 (poster)

Abstract
MR image reconstruction from undersampled k-space can be improved by nonlinear denoising estimators since they incorporate statistical prior knowledge about image sparsity. Reconstruction quality depends crucially on the undersampling design (k-space trajectory), in a manner complicated by the nonlinear and signal-dependent characteristics of these methods. We propose an algorithm to assess and optimize k-space trajectories for sparse MRI reconstruction, based on Bayesian experimental design, which is scaled up to full MR images by a novel variational relaxation to iteratively reweighted FFT or gridding computations. Designs are built sequentially by adding phase encodes predicted to be most informative, given the combination of previous measurements with image prior information.

PDF Web [BibTex]

PDF Web [BibTex]


no image
MR-Based Attenuation Correction for PET/MR

Hofmann, M., Steinke, F., Bezrukov, I., Kolb, A., Aschoff, P., Lichy, M., Erb, M., Nägele, T., Brady, M., Schölkopf, B., Pichler, B.

17(260), 17th Annual Meeting of the International Society for Magnetic Resonance in Medicine (ISMRM), April 2009 (poster)

Abstract
There has recently been a growing interest in combining PET and MR. Attenuation correction (AC), which accounts for radiation attenuation properties of the tissue, is mandatory for quantitative PET. In the case of PET/MR the attenuation map needs to be determined from the MR image. This is intrinsically difficult as MR intensities are not related to the electron density information of the attenuation map. Using ultra-short echo (UTE) acquisition, atlas registration and machine learning, we present methods that allow prediction of the attenuation map based on the MR image both for brain and whole body imaging.

PDF Web [BibTex]

PDF Web [BibTex]

2008


no image
BCPy2000

Hill, N., Schreiner, T., Puzicha, C., Farquhar, J.

Workshop "Machine Learning Open-Source Software" at NIPS, December 2008 (talk)

Web [BibTex]

2008

Web [BibTex]


no image
Logistic Regression for Graph Classification

Shervashidze, N., Tsuda, K.

NIPS Workshop on "Structured Input - Structured Output" (NIPS SISO), December 2008 (talk)

Abstract
In this paper we deal with graph classification. We propose a new algorithm for performing sparse logistic regression for graphs, which is comparable in accuracy with other methods of graph classification and produces probabilistic output in addition. Sparsity is required for the reason of interpretability, which is often necessary in domains such as bioinformatics or chemoinformatics.

Web [BibTex]

Web [BibTex]


no image
New Projected Quasi-Newton Methods with Applications

Sra, S.

Microsoft Research Tech-talk, December 2008 (talk)

Abstract
Box-constrained convex optimization problems are central to several applications in a variety of fields such as statistics, psychometrics, signal processing, medical imaging, and machine learning. Two fundamental examples are the non-negative least squares (NNLS) problem and the non-negative Kullback-Leibler (NNKL) divergence minimization problem. The non-negativity constraints are usually based on an underlying physical restriction, for e.g., when dealing with applications in astronomy, tomography, statistical estimation, or image restoration, the underlying parameters represent physical quantities such as concentration, weight, intensity, or frequency counts and are therefore only interpretable with non-negative values. Several modern optimization methods can be inefficient for simple problems such as NNLS and NNKL as they are really designed to handle far more general and complex problems. In this work we develop two simple quasi-Newton methods for solving box-constrained (differentiable) convex optimization problems that utilize the well-known BFGS and limited memory BFGS updates. We position our method between projected gradient (Rosen, 1960) and projected Newton (Bertsekas, 1982) methods, and prove its convergence under a simple Armijo step-size rule. We illustrate our method by showing applications to: Image deblurring, Positron Emission Tomography (PET) image reconstruction, and Non-negative Matrix Approximation (NMA). On medium sized data we observe performance competitive to established procedures, while for larger data the results are even better.

PDF [BibTex]

PDF [BibTex]


no image
Variational Bayesian Model Selection in Linear Gaussian State-Space based Models

Chiappa, S.

International Workshop on Flexible Modelling: Smoothing and Robustness (FMSR 2008), 2008, pages: 1, November 2008 (poster)

Web [BibTex]

Web [BibTex]


no image
MR-Based PET Attenuation Correction: Initial Results for Whole Body

Hofmann, M., Steinke, F., Aschoff, P., Lichy, M., Brady, M., Schölkopf, B., Pichler, B.

Medical Imaging Conference, October 2008 (talk)

[BibTex]

[BibTex]


no image
Nonparametric Indepedence Tests: Space Partitioning and Kernel Approaches

Gretton, A., Györfi, L.

19th International Conference on Algorithmic Learning Theory (ALT08), October 2008 (talk)

PDF Web [BibTex]

PDF Web [BibTex]


no image
Data-driven goodness-of-fit tests

Langovoy, M.

2008 Barcelona Conference on Asymptotic Statistics (BAS), September 2008 (talk)

Web [BibTex]

Web [BibTex]


no image
Towards the neural basis of the flash-lag effect

Ecker, A., Berens, P., Hoenselaar, A., Subramaniyan, M., Tolias, A., Bethge, M.

International Workshop on Aspects of Adaptive Cortex Dynamics, 2008, pages: 1, September 2008 (poster)

PDF [BibTex]

PDF [BibTex]


no image
mGene: A Novel Discriminative Gene Finder

Schweikert, G., Zeller, G., Zien, A., Behr, J., Sonnenburg, S., Philips, P., Ong, C., Rätsch, G.

Worm Genomics and Systems Biology meeting, July 2008 (talk)

[BibTex]

[BibTex]


no image
Policy Learning: A Unified Perspective With Applications In Robotics

Peters, J., Kober, J., Nguyen-Tuong, D.

8th European Workshop on Reinforcement Learning for Robotics (EWRL 2008), 8, pages: 10, July 2008 (poster)

Abstract
Policy Learning approaches are among the best suited methods for high-dimensional, continuous control systems such as anthropomorphic robot arms and humanoid robots. In this paper, we show two contributions: firstly, we show a unified perspective which allows us to derive several policy learning al- gorithms from a common point of view, i.e, policy gradient algorithms, natural- gradient algorithms and EM-like policy learning. Secondly, we present several applications to both robot motor primitive learning as well as to robot control in task space. Results both from simulation and several different real robots are shown.

PDF [BibTex]

PDF [BibTex]


no image
Discovering Common Sequence Variation in Arabidopsis thaliana

Rätsch, G., Clark, R., Schweikert, G., Toomajian, C., Ossowski, S., Zeller, G., Shinn, P., Warthman, N., Hu, T., Fu, G., Hinds, D., Cheng, H., Frazer, K., Huson, D., Schölkopf, B., Nordborg, M., Ecker, J., Weigel, D., Schneeberger, K., Bohlen, A.

16th Annual International Conference Intelligent Systems for Molecular Biology (ISMB), July 2008 (talk)

Web [BibTex]

Web [BibTex]


no image
Coding Theory in Brain-Computer Interfaces

Martens, SMM.

Soria Summerschool on Computational Mathematics "Algebraic Coding Theory" (S3CM), July 2008 (talk)

Web [BibTex]

Web [BibTex]


no image
Motor Skill Learning for Cognitive Robotics

Peters, J.

6th International Cognitive Robotics Workshop (CogRob), July 2008 (talk)

Abstract
Autonomous robots that can assist humans in situations of daily life have been a long standing vision of robotics, artificial intelligence, and cognitive sciences. A first step towards this goal is to create robots that can learn tasks triggered by environmental context or higher level instruction. However, learning techniques have yet to live up to this promise as only few methods manage to scale to high-dimensional manipulator or humanoid robots. In this tutorial, we give a general overview on motor skill learning for cognitive robotics using research at ATR, USC, CMU and Max-Planck in order to illustrate the problems in motor skill learning. For doing so, we discuss task-appropriate representations and algorithms for learning robot motor skills. Among the topics are the learning basic movements or motor primitives by imitation and reinforcement learning, learning rhytmic and discrete movements, fast regression methods for learning inverse dynamics and setups for learning task-space policies. Examples on various robots, e.g., SARCOS DB, the SARCOS Master Arm, BDI Little Dog and a Barrett WAM, are shown and include Ball-in-a-Cup, T-Ball, Juggling, Devil-Sticking, Operational Space Control and many others.

Web [BibTex]

Web [BibTex]


no image
Reinforcement Learning of Perceptual Coupling for Motor Primitives

Kober, J., Peters, J.

8th European Workshop on Reinforcement Learning for Robotics (EWRL 2008), 8, pages: 16, July 2008 (poster)

Abstract
Reinforcement learning is a natural choice for the learning of complex motor tasks by reward-related self-improvement. As the space of movements is high-dimensional and continuous, a policy parametrization is needed which can be used in this context. Traditional motor primitive approaches deal largely with open-loop policies which can only deal with small perturbations. In this paper, we present a new type of motor primitive policies which serve as closed-loop policies together with an appropriate learning algorithm. Our new motor primitives are an augmented version version of the dynamic systems motor primitives that incorporates perceptual coupling to external variables. We show that these motor primitives can perform complex tasks such a Ball-in-a-Cup or Kendama task even with large variances in the initial conditions where a human would hardly be able to learn this task. We initialize the open-loop policies by imitation learning and the perceptual coupling with a handcrafted solution. We first improve the open-loop policies and subsequently the perceptual coupling using a novel reinforcement learning method which is particularly well-suited for motor primitives.

PDF [BibTex]

PDF [BibTex]


no image
Painless Embeddings of Distributions: the Function Space View (Part 1)

Fukumizu, K., Gretton, A., Smola, A.

25th International Conference on Machine Learning (ICML), July 2008 (talk)

Abstract
This tutorial will give an introduction to the recent understanding and methodology of the kernel method: dealing with higher order statistics by embedding painlessly random variables/probability distributions. In the early days of kernel machines research, the "kernel trick" was considered a useful way of constructing nonlinear algorithms from linear ones. More recently, however, it has become clear that a potentially more far reaching use of kernels is as a linear way of dealing with higher order statistics by embedding distributions in a suitable reproducing kernel Hilbert space (RKHS). Notably, unlike the straightforward expansion of higher order moments or conventional characteristic function approach, the use of kernels or RKHS provides a painless, tractable way of embedding distributions. This line of reasoning leads naturally to the questions: what does it mean to embed a distribution in an RKHS? when is this embedding injective (and thus, when do different distributions have unique mappings)? what implications are there for learning algorithms that make use of these embeddings? This tutorial aims at answering these questions. There are a great variety of applications in machine learning and computer science, which require distribution estimation and/or comparison.

PDF Web [BibTex]

PDF Web [BibTex]


no image
Reinforcement Learning for Robotics

Peters, J.

8th European Workshop on Reinforcement Learning for Robotics (EWRL), July 2008 (talk)

Web [BibTex]

Web [BibTex]


no image
Flexible Models for Population Spike Trains

Bethge, M., Macke, J., Berens, P., Ecker, A., Tolias, A.

AREADNE 2008: Research in Encoding and Decoding of Neural Ensembles, 2, pages: 52, June 2008 (poster)

PDF [BibTex]

PDF [BibTex]


no image
Pairwise Correlations and Multineuronal Firing Patterns in the Primary Visual Cortex of the Awake, Behaving Macaque

Berens, P., Ecker, A., Subramaniyan, M., Macke, J., Hauck, P., Bethge, M., Tolias, A.

AREADNE 2008: Research in Encoding and Decoding of Neural Ensembles, 2, pages: 48, June 2008 (poster)

PDF [BibTex]

PDF [BibTex]