Search

Although an input distribution may not majorize a target distribution, it may majorize a distribution which is close to the target. Here we introduce a notion of approximate majorization. For any distribution, and given a distance $\delta$, we find the approximate distributions which majorize (are majorized by) all other distributions within the distance $\delta$. We call these the steepest and flattest approximation. This enables one to compute how close one can get to a given target distribution under a process governed by majorization. We show that the flattest and steepest approximations preserve ordering under majorization. Furthermore, we give a notion of majorization distance. This has applications ranging from thermodynamics, entanglement theory, and economics.

Due to the relative scarcity of long-term toxicity data, assessment factors for extrapolation from relatively short to chronic exposures have an important role in the risk assessment of chemicals. A recent REACH guidance document includes recommended default assessment factors that cover subacute-subchronic, subchronic-chronic, and subacute-chronic extrapolations. The recommended assessment factors are smaller than in most previous proposals, since they are calibrated to achieve central estimates (50th percentile of the target distribution) rather than a higher percentile such as the 95th, as has been more common. These assessment factors are nevertheless presented as representing a "widely agreed level of conservatism", a statement that may lead to misunderstandings of what is achieved by using them in a risk assessment. Assessment factors have been based on evidence from animal studies with different designs, in particular with focus on different endpoints. Our re-analysis of experimental data shows that using mortality as an endpoint leads to smaller assessment factors than if the factors are derived from corresponding ratios for non-lethal toxicity.

European Union (EU) road safety policies include reduction in road fatalities by 50% during 2000-2010. The original territorial distribution of this target is uniform, as all the territories have to halve the number of fatalities regardless of their previous record. We propose a simple method of distributing the total effort required to reach the EU target of halving fatalities in a territory in such a way that those areas with a higher proportion of fatalities (relative to their populations) have the highest targets and the sum of all of the areas is the 50% reduction. The distribution function we use here is based on an inverse logarithmic function selected from among several alternative functions analyzed in an initial study. This use of weighted distribution functions has been applied since 2000 by the EU in other policies, such as the use of renewable energies. We applied the proposed distribution function to two territorial aggregation levels in the EUROSTAT Nomenclature of Territorial Units for Statistics (NUTS): NUTS0 for EU-15 and EU-25 Member States, and NUTS3 for the 50 Spanish provinces, comparing the new and old targets with the real achievements for the 2000-2006 period, and new and old targets for the 2000-2010 period. This is a simple proposal for modification of target distribution that can be further improved using other parameters, such as road or weather conditions.

The minimum averaged mean squared error nonparametric adaptive weights use data from m possibly different populations to infer about one population of interest. The definition of these weights is based on the properties of the empirical distribution function. We use the Kaplan-Meier estimate to let the weights accommodate right-censored data and use them to define the weighted Kaplan-Meier estimate. The proposed estimate is smoother than the usual Kaplan-Meier estimate and converges uniformly in probability to the target distribution. Simulations show that the performances of the weighted Kaplan-Meier estimate on finite samples exceed that of the usual Kaplan-Meier estimate. A case study is also presented.

The objective of this study was to test harvester group-guiding methods by comparing how the total log output distribution of independent harvesters differs from the total log output distribution of group-guided harvesters. In this simulation study four harvesters worked in their own stands in the same region for an identical target, given by a sawmill. Group-guiding was done by utilizing other harvesters’ bucking outcomes to fulfil the target log distribution better. Harvesters were combined to form a group in an adapting phase where a new price list was formed. For group-guiding, four different price list adapting methods were developed. There were five different simulations: four with adaptation and one reference simulation without adaptation. Apportionment degree and log/pulpwood proportions were calculated to compare the difference between the methods and reference simulation. With group-guiding, by adapting the price list harvesters reached the target distribution better than working indepe...

The U.S. Customs and Border Protection reports that counterfeit goods seizures were up 83% in 2006. While a plethora of anti-counterfeiting strategies target distribution channels, international organizations, pirates, and company-based initiatives, few reports debate the effectiveness of these distinct anti-counterfeiting tactics to curb the problem. For this study we conducted in-depth interviews with United States managers to gauge the efficacy of various anti-counterfeiting tactics to preserve intellectual property rights. The results indicate that corporate managers find the practice of encouraging distributors to notify the manufacturer about counterfeits, as well as educating both employees and channel members about the counterfeit problem, to be some of the most effective ways to fight pirates. However, the managers report many other tactics are futile, including providing financial incentives for distributors to reject counterfeits and stressing the harmful effects of fake goods in advertising. We recommend a specific program that firms can employ to deter counterfeiting, including managing the registration of all trademarks and patents in key markets, establishing a company-based enforcement team, monitoring the growth of fakes through a central information repository, developing a muti-pronged action plan, and preparing to fight pirates through investigative work in conjunction with local law enforcement.

A weighted graph G is a pair (V, ℰ) containing vertex set V and edge set ℰ, where each edge e ∈ ℰ is associated with a weight We . A subgraph of G is a forest if it has no cycles. All forests on the graph G form a probability space, where the probability of each forest is proportional to the product of the weights of its edges. This paper aims to simulate forests exactly from the target distribution. Methods based on coupling from the past (CFTP) and rejection sampling are presented. Comparisons of these methods are given theoretically and via simulation.

This article proposes an agent negotiation model for target distribution across a set of geographically dispersed sensors. The key idea is to consider sensors as autonomous agents that negotiate ov ...

Belief propagation (BP) is the calculation method which enables us to obtain the marginal probabilities with a tractable computational cost. BP is known to provide true marginal probabilities when the graph describing the target distribution has a tree structure, while do approximate marginal probabilities when the graph has loops. The accuracy of loopy belief propagation (LBP) has been studied. In this paper, we focus on applying LBP to a multi-dimensional Gaussian distribution and analytically show how accurate LBP is for some cases.

В статье рассматривается решение актуальной научной проблемы уничтожения групповых воздушных целей, направленной на увеличение показателя эффективности улучшаемой системы наведения, а значит, характеристики её свойства. Обозначено решение в виде предлагаемого авторами двухэтапного сеточного метода коллективного наведения и управляемого целераспределения группы ракет в интересах эффективной атаки групповой воздушной цели, оптимального по критерию максимума вероятности её уничтожения – уничтожения всех её элементов The article discusses solving actual scientific problem destruction of group destruction of group air targets aimed at increasing the efficiency index is improved guidance system, which means that the characteristics of its properties. Denotes the solution proposed by the authors in the form of a twophase grid method of collective guidance and target distribution groups managed in the interests of effective missile attack group air targets, optimal criterion of maximum propability of its destruction – the destruction of all of its elements

We propose a simple yet effective iterative scheme that allows us to employ the well-tempered distribution as a target distribution for the collective variables in our recently introduced variational approach to enhanced sampling and free energy calculations [ Valsson and Parrinello Phys. Rev. Lett. 2014 , 113 , 090601 ]. The performance of the scheme is evaluated for the three-dimensional free energy surface of alanine tetrapeptide where the convergence can be rather poor when employing the uniform target distribution. Using the well-tempered target distribution on the other hand results in a significant improvement in convergence. The results observed in this paper indicate that the well-tempered distribution is in most cases the preferred and recommended choice for the target distribution in the variational approach.

Aiming at the performance degrade of traditional Gaussian cognitive radar (CR) system for Gaussian target, this paper extended the Gaussian target response to arbitrary non-Gaussian target distribution. In this paper, the CR multiple hypothesis classification algorithm was used for non-Gaussian targets, and the sparse spectrum of correlated narrowband target responses was utilized. In addition, we simulated the multiple non-Gaussian distribution targets, and the simulation results demonstrate that the non-Gaussian target classification algorithm is very effective.

Background: Activity cliffs (ACs) are formed by structurally analogous compounds with large potency variations. In medicinal chemistry, ACs are focal points of structure–activity relationship (SAR) analysis and typically studied on the basis of individual compound series. Results: ACs have been investigated on a large scale by analyzing active compounds from medicinal chemistry sources covering current pharmaceutical targets. Growth in AC populations has been determined and AC distributions have been analyzed. Conclusion: Over the past few years, the number of available ACs has essentially doubled, which substantially expands the knowledge base for SAR exploration. However, the AC frequency, target distribution, and proportion of AC compounds have remained constant, thus providing a consistent picture of AC formation across bioactive compounds.

Because the performance of traditional Gaussian cognitive radar(CR) system for target response varying from Gaussian will be degraded, this paper extended the Gaussian target response to arbitrary non-Gaussian target distribution. In this paper, the CR multiple hypothesis classification algorithm was used for non-Gaussian targets, and the sparse spectrum of correlated narrowband target responses was utilized. In addition, we simulated the multiple nonGaussian distribution targets, and the simulation results demonstrate the non-Gaussian target classification algorithm is very effective.

As traditional two-parameter constant false alarm rate (CFAR) target detection algorithms in SAR images ignore target distribution, their performances are not the best or near best. As the resolution of SAR images increases, small targets present more pixels in SAR images. So the target distribution is of much significance. Distribution-based CFAR detection algorithm is presented. We unite the pixels around the test cell, and estimate the distribution of test cell by them. Generalized Likelihood Ratio Test (GLRT) is used to deduce the detectors. The performance of the distribution-based CFAR (DBCFAR) detectors is analyzed theoretically. False alarms of DBCFAR detection are fewer than those of CFAR at the same detection rate. Finally experiments are done and the results show the performance of DBCFAR is out of conventional CFAR. False alarms of DBCFAR detection are concentrated while those of CFAR detection are dispersive.

Abundance estimates of out-migrating sockeye salmon Oncorhynchus nerka smolts are used to prepare preseason forecasts of adult returns and to set escapement goals. Here we describe a method for estimating smolt flux and abundance that uses side-looking sonar. This method more efficiently covers the river cross section and is logistically easier to deploy than up-looking sonar systems. To account for the skewed vertical distribution of smolts, we used a recently developed model for correcting echo integration bias associated with nonuniform target distribution. The correction is based on adjusting the integrated beam pattern for a given distribution of targets relative to the transverse cross section of the beam. We compared the results with estimates derived from video data and modeled the effect of three vertical distributions of smolts and three transducer pitch angles. The model estimated correction factors that ranged from 0.6 to 2.6. Correction factors were greater than 1 (which indicates ne...

The use of tissue microarrays (TMAs) in the determination of novel target molecule distribution in organs is an expanding area of discovery pathology. This pilot study was carried out to assess the Chromavision automated cellular imaging system (ACIS) for quantitation of both mRNA and protein distribution in rat and dog TMAs. The targets chosen were a protein kinase, P-CIP2, for mRNA assessment and its downsteam target, peptidylglycine amidating monoxygenase (PAM), for immunohistochemistry (IHC). Oligonucleotide probes produced against P-CIP2, together with an antibody against PAM, were evaluated on rat and dog TMAs. A method for evaluation of target distribution using the ACIS was developed and involved a two-tier approach. Firstly, an initial scanning of the labelled slides identified which tissues expressed the target. Secondly, a more comprehensive analysis was made. This required operator interaction to select specific regions of interest within selected tissue cores and exclude any background labelling from the final assessment. This exacted the level of expression of P-CIP2 or PAM in different cellular populations in tissue cores. A comparative semi-quantitative analysis of the same arrays was concomitantly made by the pathologist in order to assess the relative benefits of a potentially time-consuming detailed morphological evaluation. This involved the histological identification by the pathologist of specific cell populations expressing P-CIP2 or PAM. In this study, we demonstrate the power of an image analysing system to provide quantitative data on target distribution by in situ hybridisation and IHC on normal TMAs. This methodology, together with detailed histological analysis by a pathologist, forms a guideline for future target distribution evaluation within discovery pathology.

We show that rejection sampling is inferior to the importance sampling algorithm in terms of the χ 2 distance between the proposal distribution and the target distribution. Similar conclusions are drawn for comparing rejection control with importance sampling.

This paper describes the use of an adjoint method for aerodynamic and aeroacoustic optimization. An adjoint method is used to find an optimum distribution of blowing in a convergent-divergent nozzle to achieve a desired pressure distribution in the nozzle with the goal of noise reduction. This paper describes the development of a three-dimensional adjoint method for simulations on multi-block structured grids. An example is given where a particular blowing distribution is first imposed. The resulting pressure distribution is used as the target distribution with no blowing as an initial condition. It is shown that after several cycles, a blowing distribution is obtained that provides the desired or target pressure distribution.

In the past decades simulation techniques, particularly the use of Markov chain Monte Carlo (MCMC) methods, have revolutionized statistical inference (cf. [9]). There has, however, been little impact of this revolution on the experimental design literature other than the pioneering work initiated by Peter Muller (cf. [7] and [8]) and his followers.

The classical redistribution problem aims at optimally scheduling communications when moving from an initial data distribution to a target distribution where each processor will host a subset of data items. However, modern computing platforms are equipped with a powerful interconnection switch, and the cost of a given communication is (almost) independent of the location of its sender and receiver. This leads to generalizing the redistribution problem as follows: find the optimal one-toone mapping of the subsets of data items onto the processors for which the cost of the redistribution is minimal. This paper studies the complexity of this generalized problem. We provide optimal algorithms and evaluate their gain over classical redistribution through simulations. We also show the NP-hardness of the problem to find the optimal data partition and processor permutation (defined by new subsets) that minimize the cost of redistribution followed by a simple computation kernel.

International audience; We present a new and original method to solve the domain adaptation problem using optimal transport. By searching for the best transportation plan between the probability distribution functions of a source and a target domain, a non-linear and invertible transformation of the learning samples can be estimated. Any standard machine learning method can then be applied on the transformed set, which makes our method very generic. We propose a new optimal transport algorithm that incorporates label information in the optimization: this is achieved by combining an efficient matrix scaling technique together with a majoration of a non-convex regularization term. By using the proposed optimal transport with label regularization, we obtain significant increase in performance compared to the original transport solution. The proposed algorithm is computationally efficient and effective, as illustrated by its evaluation on a toy example and a challenging real life vision dataset, against which it achieves competitive results with respect to state-of-the-art methods.

The problem of universal search and stop using an adaptive search policy is considered. When the target location is searched, the observation is distributed according to the target distribution, otherwise it is distributed according to the absence distribution. A universal sequential scheme for search and stop is proposed using only the knowledge of the absence distribution, and its asymptotic performance is analyzed. The universal test is shown to yield a vanishing error probability, and to achieve the optimal reliability when the target is present, universally for every target distribution. Consequently, it is established that the knowledge of the target distribution is only useful for improving the reliability for detecting a missing target. It is also shown that a multiplicative gain for the search reliability equal to the number of searched locations is achieved by allowing adaptivity in the search., Comment: Submitted for journal publication, December 2014

Aiming at the netted radar surveillance problem, an intelligent search strategy is put forward. Firstly the target’s probability values is given, by ESM pre-search the specified airspace surveillance, as the target distribution weights; Then based on the optimization criterion of maximizing the weighted probability of detection, the distribution function of radar-target is built and the algorithm is given; Finally, the feasibility of the method is verified through MATLAB simulation experiment. The content of this paper can provide theoretical guidance for the actual engineering application. Keywords-netted surveillance radar; weak maneuvering target; weighted searching probability; MATLAB simulation

Simulation studies are used to evaluate the impact of the distributional and correlational characteristics of traffic arrival processes on the performance of network routing elements. It is shown that synthetic traffic models that capture only the distributional or the correlational characteristics of real workloads can yield substantially optimistic predictions of queue lengths and drop rate. A new technique for generating synthetic arrival streams is proposed and evaluated. Arrival streams are generated by the widely used method of sampling from a target distribution. However, the uniform stream used in the sampling is itself derived from fractional Gaussian noise. The resulting synthetic streams are shown to have sample autocorrelation functions that are consistent with long-range dependence and to provide measurably better performance estimates than standard distribution-based and FGN-based techniques.

There are many existing test methods and data reduction techniques for statistical properties of fatigue strength at a specific fatigue life, but none of them have been validated by experiment and simulation. Validation by simulation is preferable because high cycle fatigue tests are usually very subjective, not repeatable, and time consuming. This paper compares two qualitatively different approaches for evaluation of fatigue limit distributions. The analysis is based on computer simulations. The first approach contains two methods of data reduction (by Dixon and Mood as well as that by Zhang and Kececioglu) based on simulated staircase test data. Another approach includes two methods of extrapolation from shorter fatigue lives (the ray-projected and the parallel-projected methods). The analysis in the first approach is based on the assumed target distribution of the fatigue limit, while the analysis in the second approach is dependent on the method of initial extrapolation of the target distribution to the shorter lives. Several simulation-based algorithms for the generation of fatigue limit data and low cycle fatigue data were developed based on a given statistical distribution of the fatigue limit. The simulated test results were statistically evaluated against the given distributions with a target mean and two different coefficients of variation of fatigue limits (COV (SFL)). It was concluded that for a regular COV (SFL) (say 3.3%), the parallel-projected method is the best, followed by the Dixon and Mood approach. In an extremely large COV (SFL) (say 30%), the Dixon and Mood method is recommended.

In Politis and Romano (1994) the subsampling methodology was put forth for approximating the sampling distribution (and the corresponding quantiles) of general statistics from i.i.d. and stationary data. In this note, we address the question of how well the subsampling distribution approximates the tail of the target distribution. In the regular setting of the sample mean of an m m -dependent sequence we show a moderate deviation property of the subsampling distribution.

Summary A stand generation (or initialization) procedure is designed to furnish morphologically plausible model trees for process-based projection. The steps of the initialization are: (i) the generation of the locations of model trees and the tessella tion of tree areas; (ii) the sampling of diameters from a target distribution and assignments of those diameters to model-tree locations; (iii) the calculation of the height of each model tree from its assigned diameter and the distances to its neighbors; (iv) the calculation of the crown length of each model tree from its height and distances to its neighbors; and (v) the recalculation of diameter from height and crown length. Components of dry matter are calculated from the model-tree dimensions and pipe-model theory. Process-based projection with the AMORPHYS model is discussed.

We consider the Gibbs sampler as a tool for generating an absolutely continuous probability measure ≥ on Rd. When an appropriate irreducibility condition is satisfied, the Gibbs Markov chain (Xn;n ≥ 0) converges in total variation to its target distribution ≥. Sufficient conditions for geometric convergence have been given by various authors. Here we illustrate, by means of simple examples, how slow the convergence can be. In particular, we show that given a sequence of positive numbers decreasing to zero, say (bn;n ≥ 1), one can construct an absolutely continuous probability measure ≥ on Rd which is such that the total variation distance between ≥ and the distribution of Xn, converges to 0 at a rate slower than that of the sequence (bn;n ≥ 1). This can even be done in such a way that ≥ is the uniform distribution over a bounded connected open subset of Rd. Our results extend to hit-and-run samplers with direction distributions having supports with symmetric gaps. Nous considerons l'echantillonneur de Gibbs comme outil pour engendrer une loi absolument continue sur Rd, disons ≥. Lorsqu'une certaine condition d'irreductibilite est satisfaite, la chaǐne de Markov (Xn;n ≥ 0) associee a l'echantillonneur de Gibbs converge en variation totale vers sa loi cible ≥. Plusieurs auteurs ont donne des conditions suffisantes pour assurer que le taux de convergence soit geometrique. Dans le present travail, nous illustrons, au moyen d'exemples simples, la lenteur avec laquelle cette convergence peut avoir lieu. En particulier, nous montrons que pour toute suite de nombres positifs (bn;n ≥ 1) qui converge vers 0, il existe dans Rd un ensemble ouvert, borne et connexe, disons G, qui est tel que si ≥ est la loi uniforme sur G, et si (Xn;n ≥ 0) est la chaǐne de Markov associee a l'echantillonneur de Gibbs pour ≥, alors la distribution de Xn converge vers 0 plus lentement que la suite (bn;n ≥ 1). Nos resultats se generalised a l'echantillonneur hit-and-run lorsque celui-ci possede une distribution de direction avec un trou symetrique.

Summary We consider the optimal scaling problem for proposal distributions in Hastings-Metropolis algorithms derived from Langevin diffusions. We prove an asymptotic diffusion limit theorem and show that the relative efficiency of the algorithm can be characterized by its overall acceptance rate, independently of the target distribution. The asymptotically optimal acceptance rate is 0.574. We show that, as a function of dimension n, the complexity of the algorithm is O(n1/3), which compares favourably with the O(n) complexity of random walk Metropolis algorithms. We illustrate this comparison with some example simulations.

The number of random bits required to approximate a target distribution in terms of un-normalized informational divergence is considered. It is shown that for a variable-to-variable length encoder, this number is lower bounded by the entropy of the target distribution. A fixed-to-variable length encoder is constructed using M-type quantization and Tunstall coding. It is shown that the encoder achieves in the limit an un-normalized informational divergence of zero with the number of random bits per generated symbol equal to the entropy of the target distribution. Numerical results show that the proposed encoder significantly outperforms the optimal block-to-block encoder in the finite length regime., Comment: Essentially the ITW 2013 final version. Compared to v1, minor typos were corrected and Fig. 1 with an example variable length encoder was added

Markov Chain Monte Carlo (MCMC) is an invaluable means of inference with complicated models, and Hamiltonian Monte Carlo, in particular Riemannian Manifold Hamiltonian Monte Carlo (RMHMC), has demonstrated success in many challenging problems. Current RMHMC implementations, however, rely on a Riemannian metric that limits their application. In this paper I propose a new metric for RMHMC without these limitations and verify its success on a distribution that emulates many hierarchical and latent models.

Importance sampling uses observations from one distribution to estimate for another distribution by weighting the observations. Including the target distribution as one component of a mixture distribution bounds the weights and makes importance sampling more reliable. The usual importance-sampling estimate is a weighted average with weights that do not sum to 1. We discuss simple normalization and other, more efficient normalization methods. These innovations make importance sampling useful in a wider variety of problems. We demonstrate with a case study of oil-inventory reliability at a large utility.

Recent improvements to the planning and scheduling process of the Hubble Space Telescope will maximize scientific return over the telescope's remaining years. The return to 3-gyro operations increased the scheduling opportunities for science on HST. Improvements to the scheduling of programs through the South Atlantic Anomaly optimized efficiency. To address an RA-asymmetry in the HST target distribution, "exclusion zones" were implemented in the Cycle 20 Call for Proposals. These process improvements not only allow the continuation of groundbreaking science, but also serve as a bridge to the planning and scheduling systems of the James Webb Space Telescope.

Common randomness is an important resource in many areas such as game theory and cryptography. We discuss the general problem of common randomness amplification between two distrustful parties connected by a communication channel and sharing some initial randomness. In this setting, both parties wish to agree on a common value distributed according to a target distribution by using their initial amount of common randomness and exchanging messages. Our results show that no protocol which is secure in a composable sense can significantly amplify the entropy initially shared by the parties.

Published version of an article in the journal: Journal of Information Processing Systems. Also available from the publisher at: http://dx.doi.org/10.3745/JIPS.2012.8.2.191 - Open Access Since a social network by definition is so diverse, the problem of estimating the preferences of its users is becoming increasingly essential for personalized applications, which range from service recommender systems to the targeted advertising of services. However, unlike traditional estimation problems where the underlying target distribution is stationary; estimating a user's interests typically involves non-stationary distributions. The consequent time varying nature of the distribution to be tracked imposes stringent constraints on the "unlearning"capabilities of the estimator used. Therefore, resorting to strong estimators that converge with a probability of 1 is inefficient since they rely on the assumption that the distribution of the user's preferences is stationary. In this vein, we propose to use a family of stochastic-learning based Weak estimators for learning and tracking a user's time varying interests. Experimental results demonstrate that our proposed paradigm outperforms some of the traditional legacy approaches that represent the state-of-the-art technology.

Penetration level of distributed generation (DG) on the grid is increasing worldwide. For the connection of new DG to the distribution network, various factors are considered including technical requirements that ensure the DG does not adversely affect the operation and safety of the network. Distribution system operators (DSO) have to perform impact studies of DG but these studies need cumbersome modeling and simulating of the target distribution network whenever DG is connected and the long time required for technical review. Therefore, utility engineers need the screening methods that can straightforwardly indentify cases with no impact on distribution system or those applications requiring detail studies. This paper presents the screening process and evaluation methods with simple rule on bank level and feeder level, respectively.

Light emitting diodes are routinely used for general and specialized lighting tasks such as automotive headlights, streetlights, luminaires, and medical illuminators. Because of their small size and directionality, LEDs enable unique light collection and shaping techniques. This paper reviews strategies that can be used to shape light according to a desired target distribution given target uniformity and system compactness constraints imposed by each application. We address design techniques for LED secondary optics, including TIR lenses, freeform reflectors, and lens and mirror arrays.

The standard adaptive pursuit technique (AP) shows preference for a single operator at a time but is not able to simultaneously pursue multiple operators. We generalize AP by allowing any target distribution to be pursued for operator selection probabilities. We call this the generalized adaptive pursuit algorithm (GAPA). We show that the probability matching and multi-armed bandit strategies, with particular settings, can be integrated in the GAPA framework. We propose and experimentally test two instances of GAPA. Assuming that there are multiple useful operators, the multi-operator AP pursues them all simultaneously. The multi-layer AP is intended to scale up the pursuit algorithm to a large set of operators. To experimentally test the proposed GAPA instances, we introduce the adaptive genetic Pareto local search (aGPLS) that selects on-line genetic operators to restart the Pareto local search. We show on a bi-objective Quadratic assignment problem (bQAP) instance with a large number of facilities and high correlation that aGPLSs are the algorithms with best performance tested.

Published version of am article from the book:AI 2011: Advances in Artificial Intelligence. Also available from the publisher on SpringerLink:http://dx.doi.org/10.1007/978-3-642-25832-9_81 Since a social network, by definition, is so diverse, the problem of estimating the preferences of its users is becoming increasingly essential for personalized applications which range from service recommender systems to the targeted advertising of services. However, unlike traditional estimation problems where the underlying target distribution is stationary, estimating a user’s interests, typically, involves non-stationary distributions. The consequent time varying nature of the distribution to be tracked imposes stringent constraints on the “ unlearning ” capabilities of the estimator used. Therefore, resorting to strong estimators that converge with probability 1 is inefficient since they rely on the assumption that the distribution of the user’s preferences is stationary. In this vein, we propose to use a family of stochastic-learning based Weak estimators for learning and tracking user’s time varying interests. Experimental results demonstrate that our proposed paradigm outperforms some of the traditional legacy approaches that represent the state-of-the-art.

We present a simple, efficient, and computationally cheap sampling method for exploring an un-normalized multivariate density on ℝ(d), such as a posterior density, called the Polya tree sampler. The algorithm constructs an independent proposal based on an approximation of the target density. The approximation is built from a set of (initial) support points - data that act as parameters for the approximation - and the predictive density of a finite multivariate Polya tree. In an initial "warming-up" phase, the support points are iteratively relocated to regions of higher support under the target distribution to minimize the distance between the target distribution and the Polya tree predictive distribution. In the "sampling" phase, samples from the final approximating mixture of finite Polya trees are used as candidates which are accepted with a standard Metropolis-Hastings acceptance probability. Several illustrations are presented, including comparisons of the proposed approach to Metropolis-within-Gibbs and delayed rejection adaptive Metropolis algorithm.

Based on indoor experiment, the water quality model of non-persistent pollutants in paroxysmal pollution accidents of Yibin section of Yangtse River is established. Through fitting the diffusing concentration of improper pollution discharging and the actual monitoring value, the original probability of accidental pollution sources are obtained. And the original probability is modified by searching action, therefore the original probability is progressed to posteriori probability, which provided a scientific and objective method to determine the target distribution function of the search model of accidental pollution sources in a drainage area.

We consider the mean time to absorption by an absorbing target of a diffusive particle with the addition of a process whereby the particle is reset to its initial position with rate $r$. We consider several generalisations of the model of M. R. Evans and S. N. Majumdar (2011), Diffusion with stochastic resetting, Phys. Rev. Lett. 106, 160601: (i) a space dependent resetting rate $r(x)$ ii) resetting to a random position $z$ drawn from a resetting distribution ${\cal P}(z)$ iii) a spatial distribution for the absorbing target $P_T(x)$. As an example of (i) we show that the introduction of a non-resetting window around the initial position can reduce the mean time to absorption provided that the initial position is sufficiently far from the target. We address the problem of optimal resetting, that is, minimising the mean time to absorption for a given target distribution. For an exponentially decaying target distribution centred at the origin we show that a transition in the optimal resetting distribution occurs as the target distribution narrows., Comment: 17 pages, 2 figures, submitted to J. Phys. A: Math. Theor, abstract corrected

The Gram-Charlier and Cornish-Fisher expansions are tools often used to compute value at risk (VaR) in the context of skewed and leptokurtic return distributions. These approximations use the first four moments of the unknown target distribution to compute approximate distribution and quantile functions. A drawback of these approaches is the limited set of skewness and kurtosis pairs for which valid approximations are possible. We examine here an alternative to these methods with the Johnson [1949] system of distributions which also uses the first four moments as main inputs but is capable of accommodating all possible skewness and kurtosis pairs. Formulas for the expected shortfall are derived. The performance of the Cornish-Fisher, Gram-Charlier and Johnson approaches for computing value at risk and expected shortfall are compared and documented. The results reveal that Johnson distributions yield smaller approximation errors than the Gram-Charlier and Cornish-Fisher approaches when used with exact or estimated moments.

With the increasing availability of alpha emitters, targeted α-particle therapy has emerged as a solution of choice to treat haematological cancers and micrometastatic and minimal residual diseases. Alpha-particles are highly cytotoxic because of their high linear energy transfer (LET) and have a short range of a few cell diameters in tissue, assuring good treatment specificity. These radiologic features make conventional dosimetry less relevant for that context. Stochastic variations in the energy deposited in cell nuclei are important because of the microscopic target size, low number of α- particle traversals, and variation in LET along the α-particle track. Microdosimetry provides a conceptual framework that aims at a systematic analysis of the stochastic distribution of energy deposits in irradiated matter. The different quantities of microdosimetry and the different methods of microdosimetric calculations were described in the early eighties. Since then, numerous models have been published through the years and applied to analyse experimental data or to model realistic therapeutic situations. Major results have been an accurate description of the high toxicity of α-particles, and the description of the predominant effect of activity distribution at the cellular scale on toxicity or efficacy of potential targeted α-particle therapies. This last factor represents a major limitation to the use of microdosimetry in vivo because determination of the source - target distribution is complicated. The future contributions of microdosimetry in targeted α-particle therapy research will certainly depend on the ability to develop high-resolution detectors and on the implementation of pharmaco-kinetic models at the tumour microenvironment scale.

Phase information provide useful acoustic cue for identifying individual speakers. Speaker verification employing the instantaneous phase-related features that perform well in clean or matched noise/channel conditions degrades dramatically when encounter unexpected communication environments. These adverse effects can distort the short-term distributions of the speaker parameters. It is observed that by mapping each feature stream to a target distribution over a specific time interval, their robustness to environmental or channel mismatch is enhanced. Through speaker verification experiments on microphone and telephone data, it is observed that the proposed robust feature extraction front-end consistently reduces the equal error rate.

The research for this chapter was completed in spring 2004 [23]. Here we provide a revised and updated presentation of the original results. Since, a reproduction of the results and a fusion of the original detector with higherorder statistics has been published [228]. These recent advances can be seen as yet another iteration in the search for ever better cover models and should be considered as relevant starting points for future work. However, for the purpose of this book, it is useful to start with the foundations of model-based steganography and its targeted detection methods to show how it fits into our theory of cover models.

Learning algorithms relying on Gibbs sampling based stochastic approximations of the log-likelihood gradient have become a common way to train Restricted Boltzmann Machines (RBMs). We study three of these methods, Contrastive Divergence (CD) and its refined variants Persistent CD (PCD) and Fast PCD (FPCD). As the approximations are biased, the maximum of the log-likelihood is not necessarily obtained. Recently, it has been shown that CD, PCD, and FPCD can even lead to a steady decrease of the log-likelihood during learning. Taking artificial data sets from the literature we study these divergence effects in more detail. Our results indicate that the log-likelihood seems to diverge especially if the target distribution is difficult to learn for the RBM. The decrease of the likelihood can not be detected by an increase of the reconstruction error, which has been proposed as a stopping criterion for CD learning. Weight-decay with a carefully chosen weight-decay-parameter can prevent divergence.

This paper discusses different probability of a single missile searching target corresponding to different searching paths, then it gives the searching probability of cooperative multi-missiles corresponding to different paths. Several feasible searching paths are present here, and work theory and capturing theory of radar are introduced generally. Based on simplified models, searching probability of a single missile or multi-missiles corresponding to different paths is computed. Simulation demonstrates that different searching paths fit different target distribution density and area size, in order to obtain high searching efficiency, proper search path should be adopted based on particular circumstance.