50 resultados para Expectation Maximization
Resumo:
We provide a comprehensive overview of many recent algorithms for approximate inference in Gaussian process models for probabilistic binary classification. The relationships between several approaches are elucidated theoretically, and the properties of the different algorithms are corroborated by experimental results. We examine both 1) the quality of the predictive distributions and 2) the suitability of the different marginal likelihood approximations for model selection (selecting hyperparameters) and compare to a gold standard based on MCMC. Interestingly, some methods produce good predictive distributions although their marginal likelihood approximations are poor. Strong conclusions are drawn about the methods: The Expectation Propagation algorithm is almost always the method of choice unless the computational budget is very tight. We also extend existing methods in various ways, and provide unifying code implementing all approaches.
Resumo:
We describe a method to explore the configurational phase space of chemical systems. It is based on the nested sampling algorithm recently proposed by Skilling (AIP Conf. Proc. 2004, 395; J. Bayesian Anal. 2006, 1, 833) and allows us to explore the entire potential energy surface (PES) efficiently in an unbiased way. The algorithm has two parameters which directly control the trade-off between the resolution with which the space is explored and the computational cost. We demonstrate the use of nested sampling on Lennard-Jones (LJ) clusters. Nested sampling provides a straightforward approximation for the partition function; thus, evaluating expectation values of arbitrary smooth operators at arbitrary temperatures becomes a simple postprocessing step. Access to absolute free energies allows us to determine the temperature-density phase diagram for LJ cluster stability. Even for relatively small clusters, the efficiency gain over parallel tempering in calculating the heat capacity is an order of magnitude or more. Furthermore, by analyzing the topology of the resulting samples, we are able to visualize the PES in a new and illuminating way. We identify a discretely valued order parameter with basins and suprabasins of the PES, allowing a straightforward and unambiguous definition of macroscopic states of an atomistic system and the evaluation of the associated free energies.
Resumo:
The partially observable Markov decision process (POMDP) provides a popular framework for modelling spoken dialogue. This paper describes how the expectation propagation algorithm (EP) can be used to learn the parameters of the POMDP user model. Various special probability factors applicable to this task are presented, which allow the parameters be to learned when the structure of the dialogue is complex. No annotations, neither the true dialogue state nor the true semantics of user utterances, are required. Parameters optimised using the proposed techniques are shown to improve the performance of both offline transcription experiments as well as simulated dialogue management performance. ©2010 IEEE.
Resumo:
Sir John Egan’s 1998 report on the construction industry (Construction Task Force 1998) noted its confrontational and adversarial nature. Both the original report and its subsequent endorsement in Accelerating Change (Strategic Forum 2002) called for improved working relationships—so-called ‘integration’—within and between both design and construction aspects. In this paper, we report on our observations of on-site team meetings for a major UK project during its construction phase. We attended a series of team meetings and recorded the patterns of verbal interaction that took place within them. In reporting our findings, we have deliberately used a graphical method for presenting the results, in the expectation that this will make them more readily accessible to designers. Our diagrams of these interaction patterns have already proved to be intuitively and quickly understood, and have generated interest and discussion among both those we observed and others who have seen them. We noted that different patterns of communication occurred in different types of meetings. Specifically, in the problem-solving meeting, there was a richness of interaction that was largely missing from progress meetings and technical meetings. Team members expressed greater satisfaction with this problem-solving meeting where these enriched exchanges took place. By making comparisons between the different patterns, we are also able to explore functional roles and their interactions. From this and other published evidence, we conclude that good teamworking practices depend on a complex interplay of relations and dependencies embedded within the team.
Resumo:
This paper is in two parts and addresses two of getting more information out of the RF signal from three-dimensional (3D) mechanically-swept medical ultrasound . The first topic is the use of non-blind deconvolution improve the clarity of the data, particularly in the direction to the individual B-scans. The second topic is imaging. We present a robust and efficient approach to estimation and display of axial strain information. deconvolution, we calculate an estimate of the point-spread at each depth in the image using Field II. This is used as of an Expectation Maximisation (EM) framework in which ultrasound scatterer field is modelled as the product of (a) a smooth function and (b) a fine-grain varying function. the E step, a Wiener filter is used to estimate the scatterer based on an assumed piecewise smooth component. In the M , wavelet de-noising is used to estimate the piecewise smooth from the scatterer field. strain imaging, we use a quasi-static approach with efficient based algorithms. Our contributions lie in robust and 3D displacement tracking, point-wise quality-weighted , and a stable display that shows not only strain but an indication of the quality of the data at each point in the . This enables clinicians to see where the strain estimate is and where it is mostly noise. deconvolution, we present in-vivo images and simulations quantitative performance measures. With the blurred 3D taken as OdB, we get an improvement in signal to noise ratio 4.6dB with a Wiener filter alone, 4.36dB with the ForWaRD and S.18dB with our EM algorithm. For strain imaging show images based on 2D and 3D data and describe how full D analysis can be performed in about 20 seconds on a typical . We will also present initial results of our clinical study to explore the applications of our system in our local hospital. © 2008 IEEE.
Resumo:
In the design of high-speed low-power electrical generators for unmanned aircraft and spacecraft, maximization of specific output (power/weight) is of prime importance. Several magnetic circuit configurations (radial-field, axial-field, flux-squeezing, homopolar) have been proposed, and in this paper the relative merits of these configurations are subjected to a quantitative investigation over the speed range 10 000–100000 rev/min and power range 250 W-10 kW. The advantages of incorporating new high energy-density magnetic materials are described. Part I deals with establishing an equivalent circuit for permanent-magnet generators. For each configuration the equivalent circuit parameters are related to the physical dimensions of the generator components and an optimization procedure produces a minimum volume design at discrete output powers and operating speeds. The technique is illustrated by a quantitative comparison of the specific outputs of conventional radial-field generators with samarium cobalt and alnico magnets. In Part II the specific outputs of conventional, flux-squeezing, and claw-rotor magnetic circuit configurations are compared. The flux-squeezing configuration is shown to produce the highest specific output for small sizes whereas the conventional configuration is best at large sizes. For all sizes the claw-rotor configuration is significantly inferior. In Part III the power densities available from axial-field and flux-switching magnetic circuit configurations are maximized, over the power range 0.25-10 kW and speed range 10 000–100000 rpm, and compared to the results of Parts I & II. For the axial-field configuration the power density is always less than that of the conventional and flux-squeezing radial-field configurations. For the flux-switching generator, which is able to withstand relatively high mechanical forces in the rotor, the power density is again inferior to the radial-field types, but the difference is less apparent for small (low power, high speed) generator sizes. From the combined results it can be concluded that the flux-squeezing and conventional radial-field magnetic circuit configurations yield designs with minimum volume over the power and speed ranges considered. © 1985, IEEE. All rights reserved.
Resumo:
This paper describes a derivation of the adjoint low Mach number equations and their implementation and validation within a global mode solver. The advantage of using the low Mach number equations and their adjoints is that they are appropriate for flows with variable density, such as flames, but do not require resolution of acoustic waves. Two versions of the adjoint are implemented and assessed: a discrete-adjoint and a continuous-adjoint. The most unstable global mode calculated with the discrete-adjoint has exactly the same eigenvalue as the corresponding direct global mode but contains numerical artifacts near the inlet. The most unstable global mode calculated with the continuous-adjoint has no numerical artifacts but a slightly different eigenvalue. The eigenvalues converge, however, as the timestep reduces. Apart from the numerical artifacts, the mode shapes are very similar, which supports the expectation that they are otherwise equivalent. The continuous-adjoint requires less resolution and usually converges more quickly than the discrete-adjoint but is more challenging to implement. Finally, the direct and adjoint global modes are combined in order to calculate the wavemaker region of a low density jet. © 2011 Elsevier Inc.
Resumo:
Introducing a "Cheaper, Faster, Better" product in today's highly competitive market is a challenging target. Therefore, for organizations to improve their performance in this area, they need to adopt methods such as process modelling, risk mitigation and lean principles. Recently, several industries and researchers focused efforts on transferring the value orientation concept to other phases of the Product Life Cycle (PLC) such as Product Development (PD), after its evident success in manufacturing. In PD, value maximization, which is the main objective of lean theory, has been of particular interest as an improvement concept that can enhance process flow logistics and support decision-making. This paper presents an ongoing study of the current understanding of value thinking in PD (VPD) with a focus on value dimensions and implementation benefits. The purpose of this study is to consider the current state of knowledge regarding value thinking in PD, and to propose a definition of value and a framework for analyzing value delivery. The framework-named the Value Cycle Map (VCM)- intends to facilitate understanding of value and its delivery mechanism in the context of the PLC. We suggest the VCM could be used as a foundation for future research in value modelling and measurement in PD.
Resumo:
A number of recent scientific and engineering problems require signals to be decomposed into a product of a slowly varying positive envelope and a quickly varying carrier whose instantaneous frequency also varies slowly over time. Although signal processing provides algorithms for so-called amplitude-and frequency-demodulation (AFD), there are well known problems with all of the existing methods. Motivated by the fact that AFD is ill-posed, we approach the problem using probabilistic inference. The new approach, called probabilistic amplitude and frequency demodulation (PAFD), models instantaneous frequency using an auto-regressive generalization of the von Mises distribution, and the envelopes using Gaussian auto-regressive dynamics with a positivity constraint. A novel form of expectation propagation is used for inference. We demonstrate that although PAFD is computationally demanding, it outperforms previous approaches on synthetic and real signals in clean, noisy and missing data settings.
Resumo:
Copulas allow to learn marginal distributions separately from the multivariate dependence structure (copula) that links them together into a density function. Vine factorizations ease the learning of high-dimensional copulas by constructing a hierarchy of conditional bivariate copulas. However, to simplify inference, it is common to assume that each of these conditional bivariate copulas is independent from its conditioning variables. In this paper, we relax this assumption by discovering the latent functions that specify the shape of a conditional copula given its conditioning variables We learn these functions by following a Bayesian approach based on sparse Gaussian processes with expectation propagation for scalable, approximate inference. Experiments on real-world datasets show that, when modeling all conditional dependencies, we obtain better estimates of the underlying copula of the data.
Resumo:
This article investigates the role of the CoO6 octahedron distortion on the electronic properties and more particularly on the high value of the Seebeck coefficient in the BiCaCoO lamellar cobaltites. Our measurements provide clues indicating that the t2g orbital degeneracy lifting has to be considered to account for the observed high temperature limit of the thermopower. They also provide experimental arguments for locating the a1g and eg′ orbitals levels on the energy scale, through the compression of the octahedron. These results are in agreement with recent ab initio calculation including the electronic correlations and concluding for the inversion of these levels as compared to the expectation from the crystal field theory. © 2007 American Institute of Physics.
Resumo:
The role dopamine plays in decision-making has important theoretical, empirical and clinical implications. Here, we examined its precise contribution by exploiting the lesion deficit model afforded by Parkinson's disease. We studied patients in a two-stage reinforcement learning task, while they were ON and OFF dopamine replacement medication. Contrary to expectation, we found that dopaminergic drug state (ON or OFF) did not impact learning. Instead, the critical factor was drug state during the performance phase, with patients ON medication choosing correctly significantly more frequently than those OFF medication. This effect was independent of drug state during initial learning and appears to reflect a facilitation of generalization for learnt information. This inference is bolstered by our observation that neural activity in nucleus accumbens and ventromedial prefrontal cortex, measured during simultaneously acquired functional magnetic resonance imaging, represented learnt stimulus values during performance. This effect was expressed solely during the ON state with activity in these regions correlating with better performance. Our data indicate that dopamine modulation of nucleus accumbens and ventromedial prefrontal cortex exerts a specific effect on choice behaviour distinct from pure learning. The findings are in keeping with the substantial other evidence that certain aspects of learning are unaffected by dopamine lesions or depletion, and that dopamine plays a key role in performance that may be distinct from its role in learning.
Resumo:
Termination of a painful or unpleasant event can be rewarding. However, whether the brain treats relief in a similar way as it treats natural reward is unclear, and the neural processes that underlie its representation as a motivational goal remain poorly understood. We used fMRI (functional magnetic resonance imaging) to investigate how humans learn to generate expectations of pain relief. Using a pavlovian conditioning procedure, we show that subjects experiencing prolonged experimentally induced pain can be conditioned to predict pain relief. This proceeds in a manner consistent with contemporary reward-learning theory (average reward/loss reinforcement learning), reflected by neural activity in the amygdala and midbrain. Furthermore, these reward-like learning signals are mirrored by opposite aversion-like signals in lateral orbitofrontal cortex and anterior cingulate cortex. This dual coding has parallels to 'opponent process' theories in psychology and promotes a formal account of prediction and expectation during pain.
Resumo:
The role dopamine plays in decision-making has important theoretical, empirical and clinical implications. Here, we examined its precise contribution by exploiting the lesion deficit model afforded by Parkinson's disease. We studied patients in a two-stage reinforcement learning task, while they were ON and OFF dopamine replacement medication. Contrary to expectation, we found that dopaminergic drug state (ON or OFF) did not impact learning. Instead, the critical factor was drug state during the performance phase, with patients ON medication choosing correctly significantly more frequently than those OFF medication. This effect was independent of drug state during initial learning and appears to reflect a facilitation of generalization for learnt information. This inference is bolstered by our observation that neural activity in nucleus accumbens and ventromedial prefrontal cortex, measured during simultaneously acquired functional magnetic resonance imaging, represented learnt stimulus values during performance. This effect was expressed solely during the ON state with activity in these regions correlating with better performance. Our data indicate that dopamine modulation of nucleus accumbens and ventromedial prefrontal cortex exerts a specific effect on choice behaviour distinct from pure learning. The findings are in keeping with the substantial other evidence that certain aspects of learning are unaffected by dopamine lesions or depletion, and that dopamine plays a key role in performance that may be distinct from its role in learning. © 2012 The Author.
Resumo:
In this paper we develop a new approach to sparse principal component analysis (sparse PCA). We propose two single-unit and two block optimization formulations of the sparse PCA problem, aimed at extracting a single sparse dominant principal component of a data matrix, or more components at once, respectively. While the initial formulations involve nonconvex functions, and are therefore computationally intractable, we rewrite them into the form of an optimization program involving maximization of a convex function on a compact set. The dimension of the search space is decreased enormously if the data matrix has many more columns (variables) than rows. We then propose and analyze a simple gradient method suited for the task. It appears that our algorithm has best convergence properties in the case when either the objective function or the feasible set are strongly convex, which is the case with our single-unit formulations and can be enforced in the block case. Finally, we demonstrate numerically on a set of random and gene expression test problems that our approach outperforms existing algorithms both in quality of the obtained solution and in computational speed. © 2010 Michel Journée, Yurii Nesterov, Peter Richtárik and Rodolphe Sepulchre.