nips nips2008 nips2008-180 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Matthias Krauledat, Konrad Grzeska, Max Sagebaum, Benjamin Blankertz, Carmen Vidaurre, Klaus-Robert Müller, Michael Schröder
Abstract: Compared to invasive Brain-Computer Interfaces (BCI), non-invasive BCI systems based on Electroencephalogram (EEG) signals have not been applied successfully for precisely timed control tasks. In the present study, however, we demonstrate and report on the interaction of subjects with a real device: a pinball machine. Results of this study clearly show that fast and well-timed control well beyond chance level is possible, even though the environment is extremely rich and requires precisely timed and complex predictive behavior. Using machine learning methods for mental state decoding, BCI-based pinball control is possible within the first session without the necessity to employ lengthy subject training. The current study shows clearly that very compelling control with excellent timing and dynamics is possible for a non-invasive BCI. 1
Reference: text
sentIndex sentText sentNum sentScore
1 de Abstract Compared to invasive Brain-Computer Interfaces (BCI), non-invasive BCI systems based on Electroencephalogram (EEG) signals have not been applied successfully for precisely timed control tasks. [sent-17, score-0.316]
2 In the present study, however, we demonstrate and report on the interaction of subjects with a real device: a pinball machine. [sent-18, score-0.499]
3 Results of this study clearly show that fast and well-timed control well beyond chance level is possible, even though the environment is extremely rich and requires precisely timed and complex predictive behavior. [sent-19, score-0.278]
4 Using machine learning methods for mental state decoding, BCI-based pinball control is possible within the first session without the necessity to employ lengthy subject training. [sent-20, score-0.689]
5 The current study shows clearly that very compelling control with excellent timing and dynamics is possible for a non-invasive BCI. [sent-21, score-0.258]
6 1 Introduction Brain computer interfaces (BCI) have seen a rapid development towards faster and more userfriendly systems for thought-based control of devices such as video games, wheel chairs, robotic devices etc. [sent-22, score-0.369]
7 While a full control of even complex trajectories has become possible for invasive BCIs [1, 2, 3], non-invasive EEG-based systems have been considered hardly able to provide such high information transfer rates between man and machine [4, 5]. [sent-23, score-0.226]
8 This paper will show evidence that real-time BCI control of a machine is possible with little subject training. [sent-24, score-0.258]
9 The machine studied (a standard pinball machine, see Fig. [sent-25, score-0.36]
10 1 requires only two classes for control but a very fast and precise reaction; predictive behavior and learning are mandatory. [sent-26, score-0.197]
11 We 1 consider it a formidable platform for studying timing and dynamics of brain control in real-time interaction with a physical machine. [sent-27, score-0.406]
12 Furthermore this paradigm is well suited for future investigations of mental states during complex real-time tasks and decision-making processes. [sent-28, score-0.11]
13 Figure 1: Left: pinball machine used for the present study. [sent-29, score-0.36]
14 Compared to highly controlled and simplified lab settings, a pinball machine provides flow (according to the definition in [6]), a rich and complex feedback, acoustic and visual distractors, and a challenging behavioral task. [sent-32, score-0.36]
15 These components are well-known ingredients for engaging and immersive game environments [7]. [sent-33, score-0.135]
16 In case of the pinball machine model used in this study, this receives further evidence from the high sales figures that have made the Addams Family model the all-time popular pinball machine. [sent-34, score-0.72]
17 Given the reaction-time critical pinball game and the intrinsic delays imposed on the subjects by the BCI technology, it is very interesting to observe that subjects can manage to control and maintain the necessary timing and dynamics. [sent-35, score-0.935]
18 The prediction of upcoming game situations and behavioral adaptation to the machine and BCI constraints are necessary ingredients to master this difficult task. [sent-36, score-0.172]
19 3 briefly introduce the used motor paradigm, spatial filter methods, the experimental paradigm, the decoding and machine learning techniques used, Sec. [sent-39, score-0.164]
20 1 Background Neurophysiology Macroscopic brain activity during resting wakefulness contains distinct rhythms located over various brain areas. [sent-43, score-0.135]
21 Sensorimotor cortices show rhythmic macroscopic EEG oscillations (µ-rhythm or sensorimotor rhythm, SMR), with spectral peak energies of about 8–14 Hz (α-band) and/or 16–28 Hz (β-band) localized in the motor and somatosensory cortex ([8]). [sent-44, score-0.151]
22 A large class of EEG-based BCI systems relies on the fact that amplitude modulations of sensorimotor rhythms can be caused, e. [sent-45, score-0.077]
23 For example, the power of the µ-rhythm decreases during imagined hand movements in the corresponding representation area which is located in the contralateral sensorimotor cortex. [sent-48, score-0.151]
24 This phenomenon is called event-related desynchronization (ERD, [9, 10]), while the increase of band power is termed event-related synchronization (ERS). [sent-49, score-0.193]
25 , during motor imagery over flanking sensorimotor areas, possibly reflecting an ‘surround inhibition’ enhancing focal cortical activation, see [11, 10]. [sent-52, score-0.25]
26 The exact location and the exact frequency band of the sensorimotor rhythm is subject-specific. [sent-53, score-0.214]
27 Since variance of band-pass filtered signals is equal to band power, CSP analysis is applied to band-pass filtered signals in order to obtain an effective discrimination of mental states that are characterized by ERD/ERS effects (see above). [sent-66, score-0.235]
28 right hand motor imagery, the CSP algorithm will find two groups of spatial filters. [sent-68, score-0.133]
29 The first will show high band power during left hand motor imagery and low band power during right hand motor imagery, and the second vice versa. [sent-69, score-0.485]
30 , the ith spatial filter) maximizes the variance for class 1, and since 1 − di is near 0, it also minimizes the variance for class 2. [sent-76, score-0.089]
31 1 Experiment Paradigm Standard EEG lab experiments typically realize an environment that avoids distractions in order to have maximum control over all parameters of the experiment. [sent-84, score-0.226]
32 Since the subjects respond to a small number of artificial stimuli, a stimulus-locked averaging reveals the average characteristics of their brain response. [sent-85, score-0.144]
33 However, gaining even only partial introspection into the system states of complex physical devices and into the interaction processes between the system and the mental processes of the user requires a huge effort. [sent-91, score-0.287]
34 2 Setup In this study seven subjects played with the pinball machine. [sent-101, score-0.485]
35 One subject played successfully and enjoyed it, but was excluded from further analysis as his/her games had not been video-taped. [sent-103, score-0.243]
36 From the remaining six subjects, three managed to acquire good control, played very successfully and enjoyed this experience. [sent-104, score-0.083]
37 One subject managed to get limited control and reported to enjoy the games although some of his/her scores were close to chance. [sent-105, score-0.357]
38 The performance of these four subjects was measured in a rigorous manner. [sent-106, score-0.091]
39 The remaining two subjects could not establish reliable control and were also excluded from further analysis. [sent-107, score-0.288]
40 The experiment was organized in several stages: the calibration of the BCI system (Sec. [sent-110, score-0.081]
41 3), the fine-tuning of parameters in a simple cursor feedback paradigm (Sec. [sent-112, score-0.133]
42 4), the application of the BCI control system during pinball games (Sec. [sent-114, score-0.656]
43 EEG Amplifier / Digitizer Feedback Filter (FQ / spatial) Classifier Player Low-level controller Paddle control signal Figure 2: Schematic view of the BCI-controlled pinball machine. [sent-121, score-0.598]
44 The user’s EEG signals upon motor imagery are amplified, digitized, filtered in the frequency domain and the spatial domain by CSP. [sent-122, score-0.27]
45 The classifier output is translated by a low-level controller into paddle movements. [sent-124, score-0.218]
46 3 Calibration of the BCI system The BCI system was calibrated individually for each of the subjects (VPMa, VPks, VPzq, VPlf ) to discriminate two classes of motor imagery (left hand and right hand). [sent-126, score-0.264]
47 The calibration procedure followed a standard Berlin BCI (BBCI) paradigm based on spatial filters and oscillatory features that avoids and prevents the use of class-correlated EOG or EMG artefacts (see [29, 28] for details). [sent-127, score-0.179]
48 Visualizing the spatial filters and the resulting patterns of activity showed that EOG or EMG components were disregarded for the calibration of the BCI system. [sent-128, score-0.14]
49 For the calibration, 100 (VPMa) or 75 (VPks, VPzq, VPlf ) trials of motor imagery were collected for each class. [sent-129, score-0.173]
50 For every trial of 4–5s duration, the class of the motor imagery was indicated on a computer screen by visual cues. [sent-130, score-0.173]
51 The calibration procedure included the determination of a subject-specific frequency band for the mu-rhythm (see Sec. [sent-131, score-0.169]
52 All subjects showed a crossvalidation error below 10% on the calibration data. [sent-136, score-0.172]
53 4 Cursor feedback control by BCI The bias of the classifier, a gain factor and thresholds for an idle-class (for classifier outputs close to the decision plane) were adapted during a short control task running on a computer screen. [sent-138, score-0.436]
54 The subject had to control a horizontally moving cursor to a target on the left or right side of the screen 4 for approximately 2 minutes while fixating a cross in the center. [sent-139, score-0.31]
55 For an exhaustive study on the role of bias adaptation in BCI, especially in the context of changing from calibration to feedback, see [30, 24]. [sent-142, score-0.118]
56 5 Pinball control by BCI A real, physical pinball machine (in our study an Addams Family model) needs good control in terms of classification accuracy and timing (dynamics). [sent-144, score-0.862]
57 The subject has to learn the physical properties of the machine to play well. [sent-145, score-0.138]
58 1) can cause the ball to go into rather unpredictable directions. [sent-147, score-0.119]
59 This interaction with the pinball machine makes the game interesting and challenging. [sent-148, score-0.543]
60 Fast brain dynamics that participate in the eye-hand coordination and visual memory play an essential role to cope with these difficulties. [sent-149, score-0.083]
61 Three modifications were implemented in order to reduce the frequency of manual ball launches (1 and 3) and to increase the frequency of balls passing the paddle areas (1 and 2). [sent-152, score-0.41]
62 While the original character of the game was not changed, the modifications introduced slight simplification to conduct the experiment. [sent-153, score-0.135]
63 side limits that prevents balls from exiting without passing the paddles 2. [sent-156, score-0.205]
64 a soft central bump in front of the paddles that biases balls to pass one of the paddles rather than exiting in a perfect vertical trajectory. [sent-157, score-0.27]
65 a reduced slope of the game field (about half the original slope), that somewhat slows down the game speed. [sent-160, score-0.27]
66 During the BCI-controlled gaming (”bci” control mode), the subject sat in front of the pinball machine, hands resting on the arm rests except for short times when new balls had to be launched with the pulling lever. [sent-161, score-0.872]
67 The EEG signals recorded in the previous 500ms were translated by the BCI system into a control signal. [sent-162, score-0.273]
68 A simple low-level control mechanism was implemented in software that translated the continuous classifier output by thresholding into a three-class signal (left flipper, idle, right flipper) using the thresholds pre-determined during the cursor control (see Sec. [sent-163, score-0.525]
69 Furthermore it introduced a logic that translated a very long lasting control signal for the left or right class into a hold-and-shoot mechanism. [sent-166, score-0.276]
70 This allowed the user to catch slow balls rolling sideways down towards a paddle. [sent-167, score-0.146]
71 The user played several games of 10 to 12 balls each. [sent-168, score-0.279]
72 Performance was observed in terms of the playing time per ball, the score per game and the number of high-quality shots. [sent-169, score-0.207]
73 6 Pseudo random control mode This ”rand” control mode was incorporated into the experimental setup in order to deliver a fair performance baseline. [sent-172, score-0.578]
74 Here, the BCI system was up and running with the same settings as in the BCI-controlled pinball game, but no player was present. [sent-173, score-0.36]
75 Instead an EEG file previously recorded during the BCI-controlled pinball game was fed into the BCI system and generated the control signal for the pinball machine. [sent-174, score-1.093]
76 These signals produced the same statistics of paddle movements as in the real feedback setting. [sent-175, score-0.303]
77 But as the balls were launched at random time points, the paddle behavior was not synchronized with the ball positions. [sent-176, score-0.459]
78 Therefore, the pseudo random control mode marks 5 the chance level of the system. [sent-177, score-0.327]
79 In this mode several games of 10-12 balls each were performed. [sent-178, score-0.302]
80 7 No control mode For performance comparisons, two performance ratings (time per ball and points per game) were also taken for a series of balls that were launched without any paddle movements (”none” control mode). [sent-181, score-1.06]
81 4 Results As video recordings have been available for the four subjects, a detailed analysis of the game performances was possible. [sent-182, score-0.164]
82 The analysis compares three different scoring measures for BCI control (bbci), pseudo-random control (rand) and no control (none) and shows the histogram of high-quality shots per ball. [sent-185, score-0.725]
83 ball duration (median) is significantly higher for the BCI-controlled gaming (average of 15s over 81 balls) than for the pseudo-random control (average of 8s over 112 balls). [sent-187, score-0.422]
84 The increased average ball duration under BCI control is caused by the larger number of highquality shots per ball. [sent-191, score-0.491]
85 While in pseudo-random control only 7% of the balls scored more than one high-quality shot per ball, this rate raises drastically to 45% for the BCI control of subject VPMa. [sent-192, score-0.645]
86 A comparison of the game scores for 10 games of BCI control and 10 games of pseudo-random control shows, that these differ even stronger due to the nonlinear characteristic of the score. [sent-193, score-0.727]
87 Again, BCI control is significantly superior to the pseudo random control. [sent-198, score-0.235]
88 The difference in normalized histograms between BCI control and pseudo random control reveals, that even for the pooled data BCI-controlled games more often have a larger number of high-quality shots. [sent-199, score-0.574]
89 Not surprisingly, the BCI-controlled games showed a number of paddle movements in moments, when no ball was in the vicinity of the paddles. [sent-200, score-0.441]
90 As pseudo-random control mode was able to gain significantly better results than no control at all (see e. [sent-202, score-0.486]
91 In order to study this issue, the pseudo-random control was based on an EEG file, which had been previously recorded during the BCI-controlled gaming, the dynamics of the paddle movements was identical during both of these control modes. [sent-206, score-0.617]
92 Under these very similar conditions, the higher scores of the BCI control must be credited to the control ability of the BCI user, especially to the precise timing of a large number of paddle shots. [sent-207, score-0.635]
93 A video of the gaming performance which provides an impression of the astonishing level of timing and dynamical control – much better than the figures can show – is available under http://www. [sent-208, score-0.352]
94 Learning curves and traces of adaptation on the subject side, the use of error potentials as well as emerging subject specific strategy differences and many other exciting question must remain untouched in this first study. [sent-217, score-0.159]
95 Emotion, surprise and other mental states or cognitive processes that play an important role in such complex real-time paradigms still await their consideration in future studies. [sent-218, score-0.101]
96 Acknowledgments We thank Brain Products GmbH for funding and for help with the preparation of the pinball machine. [sent-219, score-0.36]
97 Learning to control a brain-machine interface for reaching and grasping by primates. [sent-241, score-0.232]
98 Neuronal ensemble control of prosthetic devices by a human with tetraplegia. [sent-270, score-0.25]
99 Mu rhythm (de)synchronization and o EEG single-trial classification of different motor imagery tasks. [sent-328, score-0.222]
100 Machine learning for real-time single-trial EEG-analysis: From brain-computer interfacing to mental state monitoring. [sent-423, score-0.132]
wordName wordTfidf (topN-words)
[('bci', 0.475), ('pinball', 0.36), ('control', 0.197), ('eeg', 0.188), ('paddle', 0.18), ('csp', 0.157), ('krauledat', 0.157), ('guido', 0.143), ('benjamin', 0.139), ('game', 0.135), ('berlin', 0.135), ('rand', 0.131), ('blankertz', 0.131), ('dornhege', 0.129), ('ball', 0.119), ('bbci', 0.114), ('balls', 0.111), ('matthias', 0.108), ('imagery', 0.099), ('games', 0.099), ('shots', 0.098), ('mode', 0.092), ('subjects', 0.091), ('band', 0.088), ('vpma', 0.082), ('calibration', 0.081), ('sensorimotor', 0.077), ('motor', 0.074), ('lters', 0.073), ('mental', 0.071), ('gabriel', 0.065), ('curio', 0.065), ('eng', 0.065), ('gaming', 0.065), ('paddles', 0.065), ('timing', 0.061), ('interfacing', 0.061), ('subject', 0.061), ('spatial', 0.059), ('ller', 0.055), ('none', 0.054), ('brain', 0.053), ('devices', 0.053), ('timed', 0.052), ('cursor', 0.052), ('enjoyed', 0.049), ('launched', 0.049), ('rhythm', 0.049), ('tangermann', 0.049), ('vpks', 0.049), ('vplf', 0.049), ('vpzq', 0.049), ('interaction', 0.048), ('germany', 0.048), ('laboratory', 0.047), ('physical', 0.047), ('shot', 0.043), ('clin', 0.043), ('desynchronization', 0.043), ('movements', 0.043), ('histograms', 0.043), ('ltered', 0.042), ('feedback', 0.042), ('signal', 0.041), ('duration', 0.041), ('technology', 0.041), ('paradigm', 0.039), ('pseudo', 0.038), ('translated', 0.038), ('signals', 0.038), ('adaptation', 0.037), ('interfaces', 0.037), ('trans', 0.037), ('per', 0.036), ('interface', 0.035), ('user', 0.035), ('played', 0.034), ('addams', 0.033), ('biomed', 0.033), ('emg', 0.033), ('introspection', 0.033), ('ipper', 0.033), ('klausrobert', 0.033), ('lemm', 0.033), ('losch', 0.033), ('notches', 0.033), ('xcsp', 0.033), ('decoding', 0.031), ('power', 0.031), ('synchronization', 0.031), ('neurophysiol', 0.031), ('play', 0.03), ('di', 0.03), ('environment', 0.029), ('video', 0.029), ('resting', 0.029), ('exiting', 0.029), ('eog', 0.029), ('invasive', 0.029), ('rajesh', 0.029)]
simIndex simValue paperId paperTitle
same-paper 1 0.9999994 180 nips-2008-Playing Pinball with non-invasive BCI
Author: Matthias Krauledat, Konrad Grzeska, Max Sagebaum, Benjamin Blankertz, Carmen Vidaurre, Klaus-Robert Müller, Michael Schröder
Abstract: Compared to invasive Brain-Computer Interfaces (BCI), non-invasive BCI systems based on Electroencephalogram (EEG) signals have not been applied successfully for precisely timed control tasks. In the present study, however, we demonstrate and report on the interaction of subjects with a real device: a pinball machine. Results of this study clearly show that fast and well-timed control well beyond chance level is possible, even though the environment is extremely rich and requires precisely timed and complex predictive behavior. Using machine learning methods for mental state decoding, BCI-based pinball control is possible within the first session without the necessity to employ lengthy subject training. The current study shows clearly that very compelling control with excellent timing and dynamics is possible for a non-invasive BCI. 1
2 0.19383068 243 nips-2008-Understanding Brain Connectivity Patterns during Motor Imagery for Brain-Computer Interfacing
Author: Moritz Grosse-wentrup
Abstract: EEG connectivity measures could provide a new type of feature space for inferring a subject’s intention in Brain-Computer Interfaces (BCIs). However, very little is known on EEG connectivity patterns for BCIs. In this study, EEG connectivity during motor imagery (MI) of the left and right is investigated in a broad frequency range across the whole scalp by combining Beamforming with Transfer Entropy and taking into account possible volume conduction effects. Observed connectivity patterns indicate that modulation intentionally induced by MI is strongest in the γ-band, i.e., above 35 Hz. Furthermore, modulation between MI and rest is found to be more pronounced than between MI of different hands. This is in contrast to results on MI obtained with bandpower features, and might provide an explanation for the so far only moderate success of connectivity features in BCIs. It is concluded that future studies on connectivity based BCIs should focus on high frequency bands and consider experimental paradigms that maximally vary cognitive demands between conditions. 1
3 0.10220225 75 nips-2008-Estimating vector fields using sparse basis field expansions
Author: Stefan Haufe, Vadim V. Nikulin, Andreas Ziehe, Klaus-Robert Müller, Guido Nolte
Abstract: We introduce a novel framework for estimating vector fields using sparse basis field expansions (S-FLEX). The notion of basis fields, which are an extension of scalar basis functions, arises naturally in our framework from a rotational invariance requirement. We consider a regression setting as well as inverse problems. All variants discussed lead to second-order cone programming formulations. While our framework is generally applicable to any type of vector field, we focus in this paper on applying it to solving the EEG/MEG inverse problem. It is shown that significantly more precise and neurophysiologically more plausible location and shape estimates of cerebral current sources from EEG/MEG measurements become possible with our method when comparing to the state-of-the-art. 1
4 0.096483544 67 nips-2008-Effects of Stimulus Type and of Error-Correcting Code Design on BCI Speller Performance
Author: Jeremy Hill, Jason Farquhar, Suzanna Martens, Felix Biessmann, Bernhard Schölkopf
Abstract: From an information-theoretic perspective, a noisy transmission system such as a visual Brain-Computer Interface (BCI) speller could benefit from the use of errorcorrecting codes. However, optimizing the code solely according to the maximal minimum-Hamming-distance criterion tends to lead to an overall increase in target frequency of target stimuli, and hence a significantly reduced average target-to-target interval (TTI), leading to difficulties in classifying the individual event-related potentials (ERPs) due to overlap and refractory effects. Clearly any change to the stimulus setup must also respect the possible psychophysiological consequences. Here we report new EEG data from experiments in which we explore stimulus types and codebooks in a within-subject design, finding an interaction between the two factors. Our data demonstrate that the traditional, rowcolumn code has particular spatial properties that lead to better performance than one would expect from its TTIs and Hamming-distances alone, but nonetheless error-correcting codes can improve performance provided the right stimulus type is used. 1
5 0.08585529 244 nips-2008-Unifying the Sensory and Motor Components of Sensorimotor Adaptation
Author: Adrian Haith, Carl P. Jackson, R. C. Miall, Sethu Vijayakumar
Abstract: Adaptation of visually guided reaching movements in novel visuomotor environments (e.g. wearing prism goggles) comprises not only motor adaptation but also substantial sensory adaptation, corresponding to shifts in the perceived spatial location of visual and proprioceptive cues. Previous computational models of the sensory component of visuomotor adaptation have assumed that it is driven purely by the discrepancy introduced between visual and proprioceptive estimates of hand position and is independent of any motor component of adaptation. We instead propose a unified model in which sensory and motor adaptation are jointly driven by optimal Bayesian estimation of the sensory and motor contributions to perceived errors. Our model is able to account for patterns of performance errors during visuomotor adaptation as well as the subsequent perceptual aftereffects. This unified model also makes the surprising prediction that force field adaptation will elicit similar perceptual shifts, even though there is never any discrepancy between visual and proprioceptive observations. We confirm this prediction with an experiment. 1
6 0.075534016 231 nips-2008-Temporal Dynamics of Cognitive Control
7 0.067108296 110 nips-2008-Kernel-ARMA for Hand Tracking and Brain-Machine interfacing During 3D Motor Control
8 0.06657812 111 nips-2008-Kernel Change-point Analysis
9 0.062871635 181 nips-2008-Policy Search for Motor Primitives in Robotics
10 0.059791569 247 nips-2008-Using Bayesian Dynamical Systems for Motion Template Libraries
11 0.058011163 211 nips-2008-Simple Local Models for Complex Dynamical Systems
12 0.05615025 33 nips-2008-Bayesian Model of Behaviour in Economic Games
13 0.055720363 187 nips-2008-Psychiatry: Insights into depression through normative decision-making models
14 0.055450361 74 nips-2008-Estimating the Location and Orientation of Complex, Correlated Neural Activity using MEG
15 0.053384099 232 nips-2008-The Conjoint Effect of Divisive Normalization and Orientation Selectivity on Redundancy Reduction
16 0.046771266 206 nips-2008-Sequential effects: Superstition or rational behavior?
17 0.044221982 240 nips-2008-Tracking Changing Stimuli in Continuous Attractor Neural Networks
18 0.043367594 71 nips-2008-Efficient Sampling for Gaussian Process Inference using Control Variables
19 0.042163167 121 nips-2008-Learning to Use Working Memory in Partially Observable Environments through Dopaminergic Reinforcement
20 0.041924927 65 nips-2008-Domain Adaptation with Multiple Sources
topicId topicWeight
[(0, -0.124), (1, 0.059), (2, 0.094), (3, 0.006), (4, 0.018), (5, 0.034), (6, -0.034), (7, 0.033), (8, 0.113), (9, 0.064), (10, 0.017), (11, 0.061), (12, -0.091), (13, 0.094), (14, -0.015), (15, 0.08), (16, 0.024), (17, 0.119), (18, -0.085), (19, 0.03), (20, 0.176), (21, 0.011), (22, 0.054), (23, 0.084), (24, -0.001), (25, -0.046), (26, -0.001), (27, 0.153), (28, -0.077), (29, -0.175), (30, 0.128), (31, -0.029), (32, 0.121), (33, -0.017), (34, -0.095), (35, 0.023), (36, 0.161), (37, 0.003), (38, 0.037), (39, -0.031), (40, -0.091), (41, 0.016), (42, -0.017), (43, -0.059), (44, 0.031), (45, 0.059), (46, 0.048), (47, -0.022), (48, -0.127), (49, -0.048)]
simIndex simValue paperId paperTitle
same-paper 1 0.95916033 180 nips-2008-Playing Pinball with non-invasive BCI
Author: Matthias Krauledat, Konrad Grzeska, Max Sagebaum, Benjamin Blankertz, Carmen Vidaurre, Klaus-Robert Müller, Michael Schröder
Abstract: Compared to invasive Brain-Computer Interfaces (BCI), non-invasive BCI systems based on Electroencephalogram (EEG) signals have not been applied successfully for precisely timed control tasks. In the present study, however, we demonstrate and report on the interaction of subjects with a real device: a pinball machine. Results of this study clearly show that fast and well-timed control well beyond chance level is possible, even though the environment is extremely rich and requires precisely timed and complex predictive behavior. Using machine learning methods for mental state decoding, BCI-based pinball control is possible within the first session without the necessity to employ lengthy subject training. The current study shows clearly that very compelling control with excellent timing and dynamics is possible for a non-invasive BCI. 1
2 0.88060129 243 nips-2008-Understanding Brain Connectivity Patterns during Motor Imagery for Brain-Computer Interfacing
Author: Moritz Grosse-wentrup
Abstract: EEG connectivity measures could provide a new type of feature space for inferring a subject’s intention in Brain-Computer Interfaces (BCIs). However, very little is known on EEG connectivity patterns for BCIs. In this study, EEG connectivity during motor imagery (MI) of the left and right is investigated in a broad frequency range across the whole scalp by combining Beamforming with Transfer Entropy and taking into account possible volume conduction effects. Observed connectivity patterns indicate that modulation intentionally induced by MI is strongest in the γ-band, i.e., above 35 Hz. Furthermore, modulation between MI and rest is found to be more pronounced than between MI of different hands. This is in contrast to results on MI obtained with bandpower features, and might provide an explanation for the so far only moderate success of connectivity features in BCIs. It is concluded that future studies on connectivity based BCIs should focus on high frequency bands and consider experimental paradigms that maximally vary cognitive demands between conditions. 1
3 0.58150917 75 nips-2008-Estimating vector fields using sparse basis field expansions
Author: Stefan Haufe, Vadim V. Nikulin, Andreas Ziehe, Klaus-Robert Müller, Guido Nolte
Abstract: We introduce a novel framework for estimating vector fields using sparse basis field expansions (S-FLEX). The notion of basis fields, which are an extension of scalar basis functions, arises naturally in our framework from a rotational invariance requirement. We consider a regression setting as well as inverse problems. All variants discussed lead to second-order cone programming formulations. While our framework is generally applicable to any type of vector field, we focus in this paper on applying it to solving the EEG/MEG inverse problem. It is shown that significantly more precise and neurophysiologically more plausible location and shape estimates of cerebral current sources from EEG/MEG measurements become possible with our method when comparing to the state-of-the-art. 1
4 0.5088678 110 nips-2008-Kernel-ARMA for Hand Tracking and Brain-Machine interfacing During 3D Motor Control
Author: Lavi Shpigelman, Hagai Lalazar, Eilon Vaadia
Abstract: Using machine learning algorithms to decode intended behavior from neural activity serves a dual purpose. First, these tools allow patients to interact with their environment through a Brain-Machine Interface (BMI). Second, analyzing the characteristics of such methods can reveal the relative significance of various features of neural activity, task stimuli, and behavior. In this study we adapted, implemented and tested a machine learning method called Kernel Auto-Regressive Moving Average (KARMA), for the task of inferring movements from neural activity in primary motor cortex. Our version of this algorithm is used in an online learning setting and is updated after a sequence of inferred movements is completed. We first used it to track real hand movements executed by a monkey in a standard 3D reaching task. We then applied it in a closed-loop BMI setting to infer intended movement, while the monkey’s arms were comfortably restrained, thus performing the task using the BMI alone. KARMA is a recurrent method that learns a nonlinear model of output dynamics. It uses similarity functions (termed kernels) to compare between inputs. These kernels can be structured to incorporate domain knowledge into the method. We compare KARMA to various state-of-the-art methods by evaluating tracking performance and present results from the KARMA based BMI experiments. 1
5 0.48100328 244 nips-2008-Unifying the Sensory and Motor Components of Sensorimotor Adaptation
Author: Adrian Haith, Carl P. Jackson, R. C. Miall, Sethu Vijayakumar
Abstract: Adaptation of visually guided reaching movements in novel visuomotor environments (e.g. wearing prism goggles) comprises not only motor adaptation but also substantial sensory adaptation, corresponding to shifts in the perceived spatial location of visual and proprioceptive cues. Previous computational models of the sensory component of visuomotor adaptation have assumed that it is driven purely by the discrepancy introduced between visual and proprioceptive estimates of hand position and is independent of any motor component of adaptation. We instead propose a unified model in which sensory and motor adaptation are jointly driven by optimal Bayesian estimation of the sensory and motor contributions to perceived errors. Our model is able to account for patterns of performance errors during visuomotor adaptation as well as the subsequent perceptual aftereffects. This unified model also makes the surprising prediction that force field adaptation will elicit similar perceptual shifts, even though there is never any discrepancy between visual and proprioceptive observations. We confirm this prediction with an experiment. 1
6 0.46212137 67 nips-2008-Effects of Stimulus Type and of Error-Correcting Code Design on BCI Speller Performance
7 0.43358761 187 nips-2008-Psychiatry: Insights into depression through normative decision-making models
8 0.42899445 74 nips-2008-Estimating the Location and Orientation of Complex, Correlated Neural Activity using MEG
9 0.35836247 33 nips-2008-Bayesian Model of Behaviour in Economic Games
10 0.35546818 231 nips-2008-Temporal Dynamics of Cognitive Control
11 0.31292361 211 nips-2008-Simple Local Models for Complex Dynamical Systems
12 0.31174418 30 nips-2008-Bayesian Experimental Design of Magnetic Resonance Imaging Sequences
13 0.30462751 111 nips-2008-Kernel Change-point Analysis
14 0.29674244 247 nips-2008-Using Bayesian Dynamical Systems for Motion Template Libraries
15 0.29009461 181 nips-2008-Policy Search for Motor Primitives in Robotics
16 0.28696835 121 nips-2008-Learning to Use Working Memory in Partially Observable Environments through Dopaminergic Reinforcement
17 0.27824548 186 nips-2008-Probabilistic detection of short events, with application to critical care monitoring
18 0.26735869 82 nips-2008-Fast Computation of Posterior Mode in Multi-Level Hierarchical Models
19 0.26429909 90 nips-2008-Gaussian-process factor analysis for low-dimensional single-trial analysis of neural population activity
20 0.26395497 222 nips-2008-Stress, noradrenaline, and realistic prediction of mouse behaviour using reinforcement learning
topicId topicWeight
[(6, 0.059), (7, 0.061), (12, 0.037), (15, 0.014), (28, 0.154), (57, 0.033), (59, 0.015), (63, 0.019), (71, 0.018), (77, 0.038), (78, 0.019), (81, 0.403), (83, 0.031)]
simIndex simValue paperId paperTitle
same-paper 1 0.80006444 180 nips-2008-Playing Pinball with non-invasive BCI
Author: Matthias Krauledat, Konrad Grzeska, Max Sagebaum, Benjamin Blankertz, Carmen Vidaurre, Klaus-Robert Müller, Michael Schröder
Abstract: Compared to invasive Brain-Computer Interfaces (BCI), non-invasive BCI systems based on Electroencephalogram (EEG) signals have not been applied successfully for precisely timed control tasks. In the present study, however, we demonstrate and report on the interaction of subjects with a real device: a pinball machine. Results of this study clearly show that fast and well-timed control well beyond chance level is possible, even though the environment is extremely rich and requires precisely timed and complex predictive behavior. Using machine learning methods for mental state decoding, BCI-based pinball control is possible within the first session without the necessity to employ lengthy subject training. The current study shows clearly that very compelling control with excellent timing and dynamics is possible for a non-invasive BCI. 1
2 0.59619504 201 nips-2008-Robust Near-Isometric Matching via Structured Learning of Graphical Models
Author: Alex J. Smola, Julian J. Mcauley, Tibério S. Caetano
Abstract: Models for near-rigid shape matching are typically based on distance-related features, in order to infer matches that are consistent with the isometric assumption. However, real shapes from image datasets, even when expected to be related by “almost isometric” transformations, are actually subject not only to noise but also, to some limited degree, to variations in appearance and scale. In this paper, we introduce a graphical model that parameterises appearance, distance, and angle features and we learn all of the involved parameters via structured prediction. The outcome is a model for near-rigid shape matching which is robust in the sense that it is able to capture the possibly limited but still important scale and appearance variations. Our experimental results reveal substantial improvements upon recent successful models, while maintaining similar running times. 1
3 0.58291036 226 nips-2008-Supervised Dictionary Learning
Author: Julien Mairal, Jean Ponce, Guillermo Sapiro, Andrew Zisserman, Francis R. Bach
Abstract: It is now well established that sparse signal models are well suited for restoration tasks and can be effectively learned from audio, image, and video data. Recent research has been aimed at learning discriminative sparse models instead of purely reconstructive ones. This paper proposes a new step in that direction, with a novel sparse representation for signals belonging to different classes in terms of a shared dictionary and discriminative class models. The linear version of the proposed model admits a simple probabilistic interpretation, while its most general variant admits an interpretation in terms of kernels. An optimization framework for learning all the components of the proposed model is presented, along with experimental results on standard handwritten digit and texture classification tasks. 1
4 0.45213896 75 nips-2008-Estimating vector fields using sparse basis field expansions
Author: Stefan Haufe, Vadim V. Nikulin, Andreas Ziehe, Klaus-Robert Müller, Guido Nolte
Abstract: We introduce a novel framework for estimating vector fields using sparse basis field expansions (S-FLEX). The notion of basis fields, which are an extension of scalar basis functions, arises naturally in our framework from a rotational invariance requirement. We consider a regression setting as well as inverse problems. All variants discussed lead to second-order cone programming formulations. While our framework is generally applicable to any type of vector field, we focus in this paper on applying it to solving the EEG/MEG inverse problem. It is shown that significantly more precise and neurophysiologically more plausible location and shape estimates of cerebral current sources from EEG/MEG measurements become possible with our method when comparing to the state-of-the-art. 1
5 0.44609627 246 nips-2008-Unsupervised Learning of Visual Sense Models for Polysemous Words
Author: Kate Saenko, Trevor Darrell
Abstract: Polysemy is a problem for methods that exploit image search engines to build object category models. Existing unsupervised approaches do not take word sense into consideration. We propose a new method that uses a dictionary to learn models of visual word sense from a large collection of unlabeled web data. The use of LDA to discover a latent sense space makes the model robust despite the very limited nature of dictionary definitions. The definitions are used to learn a distribution in the latent space that best represents a sense. The algorithm then uses the text surrounding image links to retrieve images with high probability of a particular dictionary sense. An object classifier is trained on the resulting sense-specific images. We evaluate our method on a dataset obtained by searching the web for polysemous words. Category classification experiments show that our dictionarybased approach outperforms baseline methods. 1
6 0.44497341 243 nips-2008-Understanding Brain Connectivity Patterns during Motor Imagery for Brain-Computer Interfacing
7 0.4406895 16 nips-2008-Adaptive Template Matching with Shift-Invariant Semi-NMF
9 0.41978252 79 nips-2008-Exploring Large Feature Spaces with Hierarchical Multiple Kernel Learning
10 0.41897497 231 nips-2008-Temporal Dynamics of Cognitive Control
11 0.41852331 196 nips-2008-Relative Margin Machines
12 0.41846603 96 nips-2008-Hebbian Learning of Bayes Optimal Decisions
13 0.41778877 4 nips-2008-A Scalable Hierarchical Distributed Language Model
14 0.41772127 106 nips-2008-Inferring rankings under constrained sensing
15 0.41731501 202 nips-2008-Robust Regression and Lasso
16 0.4173077 195 nips-2008-Regularized Policy Iteration
17 0.41730255 205 nips-2008-Semi-supervised Learning with Weakly-Related Unlabeled Data : Towards Better Text Categorization
18 0.41677156 162 nips-2008-On the Design of Loss Functions for Classification: theory, robustness to outliers, and SavageBoost
19 0.41653046 49 nips-2008-Clusters and Coarse Partitions in LP Relaxations
20 0.41637871 21 nips-2008-An Homotopy Algorithm for the Lasso with Online Observations