Results 11  20
of
446
Blind source separation of more sources than mixtures using overcomplete representations
 IEEE Sig. Proc. Lett
, 1999
"... Abstract—Empirical results were obtained for the blind source separation of more sources than mixtures using a recently proposed framework for learning overcomplete representations. This technique assumes a linear mixing model with additive noise and involves two steps: 1) learning an overcomplete r ..."
Abstract

Cited by 134 (3 self)
 Add to MetaCart
Abstract—Empirical results were obtained for the blind source separation of more sources than mixtures using a recently proposed framework for learning overcomplete representations. This technique assumes a linear mixing model with additive noise and involves two steps: 1) learning an overcomplete representation for the observed data and 2) inferring sources given a sparse prior on the coefficients. We demonstrate that three speech signals can be separated with good fidelity given only two mixtures of the three signals. Similar results were obtained with mixtures of two speech signals and one music signal. Index Terms—Blind source separation, independent component analysis, overcomplete dictionary, overcomplete representation, speech signal separation. (a) (b)
Stochastic Variational Inference
 JOURNAL OF MACHINE LEARNING RESEARCH (2013, IN PRESS)
, 2013
"... We develop stochastic variational inference, a scalable algorithm for approximating posterior distributions. We develop this technique for a large class of probabilistic models and we demonstrate it with two probabilistic topic models, latent Dirichlet allocation and the hierarchical Dirichlet proce ..."
Abstract

Cited by 131 (27 self)
 Add to MetaCart
We develop stochastic variational inference, a scalable algorithm for approximating posterior distributions. We develop this technique for a large class of probabilistic models and we demonstrate it with two probabilistic topic models, latent Dirichlet allocation and the hierarchical Dirichlet process topic model. Using stochastic variational inference, we analyze several large collections of documents: 300K articles from Nature, 1.8M articles from The New York Times, and 3.8M articles from Wikipedia. Stochastic inference can easily handle data sets of this size and outperforms traditional variational inference, which can only handle a smaller subset. (We also show that the Bayesian nonparametric topic model outperforms its parametric counterpart.) Stochastic variational inference lets us apply complex Bayesian models to massive data sets.
Mining eventrelated brain dynamics,”
 Trends in Cognitive Sciences,
, 2004
"... This article provides a new, more comprehensive view of eventrelated brain dynamics founded on an informationbased approach to modeling electroencephalographic (EEG) dynamics. Most EEG research focuses either on peaks 'evoked' in average eventrelated potentials (ERPs) or on changes &apo ..."
Abstract

Cited by 130 (21 self)
 Add to MetaCart
(Show Context)
This article provides a new, more comprehensive view of eventrelated brain dynamics founded on an informationbased approach to modeling electroencephalographic (EEG) dynamics. Most EEG research focuses either on peaks 'evoked' in average eventrelated potentials (ERPs) or on changes 'induced' in the EEG power spectrum by experimental events. Although these measures are nearly complementary, they do not fully model the eventrelated dynamics in the data, and cannot isolate the signals of the contributing cortical areas. We propose that many ERPs and other EEG features are better viewed as time/frequency perturbations of underlying field potential processes. The new approach combines independent component analysis (ICA), time/frequency analysis, and trialbytrial visualization that measures EEG source dynamics without requiring an explicit head model. Scalp EEG signals are produced by partial synchronization of neuronalscale field potentials across areas of cortex of centimetresquared scale. Although once viewed by some as a form of brain 'noise', it appears increasingly probable that this synchronization optimizes relations between spikemediated 'topdown' and 'bottomup' communication, both within and between brain areas. This optimization might have particular importance during motivated anticipation of, and attention to, meaningful events and associations and in response to their anticipated consequences [1 3]. This new view of cortical and scalprecorded field dynamics requires a new data analysis approach. Here, we suggest how a combination of signal processing and visualization methods can give a more adequate model of the spatially distributed eventrelated EEG dynamics that support cognitive events. Traditional analysis of eventrelated EEG data proceeds in one of two directions. In the timedomain approach, researchers average a set of data trials or epochs timelocked to some class of events, yielding an ERP waveform at each data channel. The frequencydomain approach averages changes in the frequency power spectrum of the whole EEG data time locked to the same events, producing a twodimensional image that we call the eventrelated spectral perturbation (ERSP; see Box 1). Neither ERP nor ERSP measures of eventrelated data fully model their dynamics. Imagine, by analogy, a snapshot of a seashore view created by averaging together a large number of snapshots taken at different times. This average snapshot would not show the waves! Similarly, ERP averaging filters out most of the EEG data, leaving only a small portion phaselocked to the timelocking events (see Box 1). The ERP and ERSP are nearly complementary. Oscillatory (ERSP) changes 'induced' by experimental events can be poorly represented in, or completely absent from the timedomain features of the ERP 'evoked' by the same events
Policy gradient methods for robotics
 In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS
, 2006
"... Abstract — The aquisition and improvement of motor skills and control policies for robotics from trial and error is of essential importance if robots should ever leave precisely prestructured environments. However, to date only few existing reinforcement learning methods have been scaled into the d ..."
Abstract

Cited by 120 (22 self)
 Add to MetaCart
(Show Context)
Abstract — The aquisition and improvement of motor skills and control policies for robotics from trial and error is of essential importance if robots should ever leave precisely prestructured environments. However, to date only few existing reinforcement learning methods have been scaled into the domains of highdimensional robots such as manipulator, legged or humanoid robots. Policy gradient methods remain one of the few exceptions and have found a variety of applications. Nevertheless, the application of such methods is not without peril if done in an uninformed manner. In this paper, we give an overview on learning with policy gradient methods for robotics with a strong focus on recent advances in the field. We outline previous applications to robotics and show how the most recently developed methods can significantly improve learning performance. Finally, we evaluate our most promising algorithm in the application of hitting a baseball with an anthropomorphic arm. I.
A robust and precise method for solving the permutation problem of frequencydomain blind source separation
 IEEE Trans. on Speech and Audio Processing 12
, 2004
"... This paper presents a robust and precise method for solving the permutation problem of frequencydomain blind source separation. It is based on two previous approaches: the direction of arrival estimation and the interfrequency correlation. We discuss the advantages and disadvantages of the two app ..."
Abstract

Cited by 116 (31 self)
 Add to MetaCart
(Show Context)
This paper presents a robust and precise method for solving the permutation problem of frequencydomain blind source separation. It is based on two previous approaches: the direction of arrival estimation and the interfrequency correlation. We discuss the advantages and disadvantages of the two approaches, and integrate them to exploit their respective advantages. We also present a closed form formula to estimate the directions of source signals from a separating matrix obtained by ICA. Experimental results show that our method solved permutation problems almost perfectly for a situation that two sources were mixed in a room whose reverberation time was 300 ms. 1.
Conditions for nonnegative independent component analysis
 IEEE Signal Processing Letters
, 2002
"... We consider the noiseless linear independent component analysis problem, in the case where the hidden sources s are nonnegative. We assume that the random variables s i s are wellgrounded in that they have a nonvanishing pdf in the (positive) neighbourhood of zero. For an orthonormal rotation y = ..."
Abstract

Cited by 96 (12 self)
 Add to MetaCart
(Show Context)
We consider the noiseless linear independent component analysis problem, in the case where the hidden sources s are nonnegative. We assume that the random variables s i s are wellgrounded in that they have a nonvanishing pdf in the (positive) neighbourhood of zero. For an orthonormal rotation y = Wx of prewhitened observations x = QAs, under certain reasonable conditions we show that y is a permutation of the s (apart from a scaling factor) if and only if y is nonnegative with probability 1. We suggest that this may enable the construction of practical learning algorithms, particularly for sparse nonnegative sources.
Natural ActorCritic
, 2007
"... In this paper, we suggest a novel reinforcement learning architecture, the Natural ActorCritic. The actor updates are achieved using stochastic policy gradients employing Amari’s natural gradient approach, while the critic obtains both the natural policy gradient and additional parameters of a valu ..."
Abstract

Cited by 95 (10 self)
 Add to MetaCart
(Show Context)
In this paper, we suggest a novel reinforcement learning architecture, the Natural ActorCritic. The actor updates are achieved using stochastic policy gradients employing Amari’s natural gradient approach, while the critic obtains both the natural policy gradient and additional parameters of a value function simultaneously by linear regression. We show that actor improvements with natural policy gradients are particularly appealing as these are independent of coordinate frame of the chosen policy representation, and can be estimated more efficiently than regular policy gradients. The critic makes use of a special basis function parameterization motivated by the policygradient compatible function approximation. We show that several wellknown reinforcement learning methods such as the original ActorCritic and Bradtke’s Linear Quadratic QLearning are in fact Natural ActorCritic algorithms. Empirical evaluations illustrate the effectiveness of our techniques in comparison to previous methods, and also demonstrate their applicability for learning control on an anthropomorphic robot arm.
Bayesian update of dialogue state: A POMDP framework for spoken dialogue systems
"... Based on the framework of partially observable Markov decision processes (POMDPs), this paper describes a practical realtime spoken dialogue system in which the underlying belief state is represented by a dynamic Bayesian Network and the policy is parameterized using a set of actiondependent basis ..."
Abstract

Cited by 79 (33 self)
 Add to MetaCart
(Show Context)
Based on the framework of partially observable Markov decision processes (POMDPs), this paper describes a practical realtime spoken dialogue system in which the underlying belief state is represented by a dynamic Bayesian Network and the policy is parameterized using a set of actiondependent basis functions. Tractable realtime Bayesian belief updating is made possible using a novel form of Loopy Belief Propagation and policy optimisation is performed using an episodic Natural Actor Critic algorithm. Details of these algorithms are provided along with evaluations of their accuracy and efficiency. The proposed POMDPbased architecture was tested using both simulations and a user trial. Both indicated that the incorporation of Bayesian belief updating significantly increases robustness to noise compared to traditional dialogue state estimation approaches. Furthermore, policy learning worked effectively and the learned policy outperformed all others on simulations. In user trials the learned policy was also competitive, although its optimality was less conclusive. Overall, the Bayesian update of dialogue state framework was shown to be a feasible and effective approach to building realworld POMDPbased dialogue systems.