Maximum entropy

The disclosure of audio-visual meeting recordings is a new challenging domain studied by several large scale research projects in Europe and the US. Automatic meeting summarization is one of the functionalities studied. In this paper we report the results of a feasibility study on a subtask, namely the summarization of meeting transcripts. A Maximum Entropy based extractive summarization system using a mix of 15 features improved the performance of a baseline system selecting all utterances longer than 10 words with 20%( ...

This paper suggests estimators of the frequencies (N. or proportions of Ndistinguishable objects contained in categories , given various types of information. We consider information in the form of exact constraints on the N., sample frequencies , and frequencies of related data. The analysis uses Bayesian methods, where the prior distribution is assumed to be a function of the crossentropy between the N. and a reference distribution. We show the relationship between our estimator and the log-linear and logit models and also present a sampling experiment to compare our proposed estimator with the iterated proportional fit ting estimator. ,.

The Boltzmann equation for transport in semiconductors is projected onto spherical harmonics in such a way that the resultant balance equations for the coefficients of the distribution function times the generalized density of states can be discretized over energy and real space by box integration. This ensures exact current continuity for the discrete equations. Spurious oscillations of the distribution function are successfully suppressed by stabilization based on a maximum entropy dissipation principle avoiding the H-transformation. The derived formulation can be used on arbitrary grids as long as box integration is possible. The new approach works not only with analytical bands but also with full-band structures in the case of holes. Results are presented for holes in bulk silicon based on a full band structure and electrons in a Si NPN BJT. For the first time the convergence of the spherical harmonics expansion is shown for a device and it is found that the quasiballistic transport in nanoscale devices requires an expansion of considerably higher order than the usual first one. The stability of the discretization is demonstrated for a range of grid spacings in the real space and bias points which produce huge gradients in the electron density and electric field. It is shown that the resultant large linear system of equations can be solved memory efficiently by the numerically robust package ILUPACK.

We have developed a new software tool, LENSVIEW, for modelling resolved gravitational lens images. Based on the LENSMEM algorithm, the software finds the best fitting lens mass model and source brightness distribution using a maximum entropy constraint. The method can be used with any point spread function or lens model. We review the algorithm and introduce some significant improvements. We also investigate and discuss issues associated with the statistical uncertainties of models and model parameters and the issues of source plane size and source pixel size.

In this work, we present a method for approximating constrained maximum entropy (ME) reconstructions of SPECT data with modifications to a block-iterative maximum a posteriori (MAP) algorithm. Maximum likelihood (ML)-based reconstruction algorithms require some form of noise smoothing. Constrained ME provides a more formal method of noise smoothing without requiring the user to select parameters. In the context of SPECT, constrained ME seeks the minimum-information image estimate among those whose projections are a given distance from the noisy measured data, with that distance determined by the magnitude of the Poisson noise. Images that meet the distance criterion are referred to as feasible images. We find that modeling of all principal degrading factors (attenuation, detector response, and scatter) in the reconstruction is critical because feasibility is not meaningful unless the projection model is as accurate as possible. Because the constrained ME solution is the same as a MAP solution for a particular value of the MAP weighting parameter, beta, the constrained ME solution can be found with a MAP algorithm if the correct value of beta is found. We show that the RBI-MAP algorithm, if used with a dynamic scheme for estimating beta, can approximate constrained ME solutions in 20 or fewer iterations. We compare results for various methods of achieving feasible images on a simulation of Tl-201 cardiac SPECT data. Results show that the RBI-MAP ME approximation provides images and quantitative estimates close to those from a slower algorithm that gives the true ME solution. Also, we find that the ME results have higher spatial resolution and greater high-frequency noise content than a feasibility-based stopping rule, feasibility-based low-pass filtering, and a quadratic Gibbs prior with beta selected according to the feasibility criterion. We conclude that fast ME approximation is possible using either RBI-MAP with the dynamic procedure or a feasibility-based stopping rule, and that such reconstructions may be particularly useful in applications where resolution is critical.

This paper considers the problem of direction-ofarrival (DOA) estimation for multiple uncorrelated plane waves incident on so-called "fully augmentable" sparse linear arrays. In situations where a decision is made on the number of existing signal sources (m m m) prior to the estimation stage, we investigate the conditions under which DOA estimation accuracy is effective (in the maximum-likelihood sense). In the case where m m m is less than the number of antenna sensors (M M M), a new approach called "MUSIC-maximum-entropy equalization" is proposed to improve DOA estimation performance in the "preasymptotic region" of finite sample size (N N N) and signal-tonoise ratio. A full-sized positive definite (p.d.) Toeplitz matrix is constructed from the M M M 2 2 2M M M direct data covariance matrix, and then, alternating projections are applied to find a p.d. Toeplitz matrix with m m m-variate signal eigensubspace ("signal subspace truncation"). When m m m M M M, Cramér-Rao bound analysis suggests that the minimal useful sample size N N N is rather large, even for arbitrarily strong signals. It is demonstrated that the well-known direct augmentation approach (DAA) cannot approach the accuracy of the corresponding Cramér-Rao bound, even asymptotically (as N N N ! ! !1 1 1) and, therefore, needs to be improved. We present a new estimation method whereby signal subspace truncation of the DAA augmented matrix is used for initialization and is followed by a local maximum-likelihood optimization routine. The accuracy of this method is demonstrated to be asymptotically optimal for the various superior scenarios (m m m M M M) presented.

... rich knowledge bases and vocabularies of clinical entities, such as the Unified Medical Language System (UMLS) (Lindberg, Humphreys, & McCray, 1993), which ... us-ing the existing MedEx tool, reported an F-score of 82.1% (Doan, Bastarache L., Klimkowski S., Denny JC, & Xu ...

The concepts of symbolic dynamics and partitioning of time series data have been used for feature extraction and anomaly detection. Although much attention has been paid to modeling of finite state machines from symbol sequences, similar efforts have not been expended for partitioning of time series data to optimally generate symbol sequences. This paper addresses this issue and proposes a partitioning method based on maximum migration of data points across cell boundaries. Various aspects of the proposed partitioning tool, such as identification of evolution characteristics of dynamical systems and adaptive selection of alphabet size, are discussed. Experimental results on an electronic circuit apparatus implementing the Duffing equation show that maximum-migration partitioning yields significant improvement over existing partitioning methods (e.g., maximum entropy partitioning) for the purpose of anomaly detection.

Using explosion source, seismic refraction data, recorded in the 1978 and 1980 Yellowstone-Snake River Plain seismic experiments, a three-dimensional inversion of differential P wave attenuation was used to assess the relative variations in Q-1 in and around the volcanically active, 45 km by 70 km, Yellowstone caldera, northwestem Wyoming. Differential attenuation was derived from spectral decay of upper crustal Pg phases, observed from six explosions and recorded at 90 temporary stations. Because of the relatively short •xme windows used to determine the spectral content, a maximum entropy technique was employed to estimate the spectra that yielded an optimally small variance. Differential P wave attenuation was calculated from least squares determinations of the spectral ratios corrected for source and path effects. The observed differential attenuation parameters were then inverted using a weighted least squares technique for a discretized, 70 km x 105 km, three-dimensional surface and upper crustal Q-1 model of the Yellowstone caldera and surrounding region. Results showed that the surface layer, to depths of 2 km within the Yellowstone caldera, is characterized by relatively high attenuation with low Q values less than 30, compared to values of 38 to 50 outside the caldera. The higher attenuation in the caldera's surface layer is thought to be associated with Quaternary lake sediments, highly altered rhyolites, and the possible influence of steam in areas of hydrothermal activity. In the crystalline upper crust, at depths of 2 km to 12 km, Q values of 40 to 70 were observed in areas of thick sedimentary fill northwest of the caldera and in areas of hydrothermal activity. Within the caldera, upper crustal attenuation generally corresponded to Q of 200 in areas that are interpreted to be associated with hot but now solidified granitic material. In comparison, relatively high attenuation, Q = 40, was observed in the upper crust of the northeastern Yellowstone caldera in an area that also corresponds to a 20% reduction in Pg velocity, a prominent negative Bouguer gravity low of-20 mGal, near Yellowstone's largest area of hydrothermal activity and near a Quatemary, volcanic resurgent dome. On the basis of constitutive models of velocity and density, this zone of high attenuation is likely a product of unusual hydrothermal activity, highly altered upper-crustal rocks, a shallow magmatic source or a highly porous, steam-saturated body. Outside the caldera, the upper crust is relatively unaffected by high temperatures and thermal alteration, and the corresponding attenuation is low with Q values greater than 200. the late Cenozoic, Yellowstone-Snake River Plain bimodal silicic-basaltic volcanic system making up the Yellowstone Plateau (Figure 1). Within the plateau, the Quaternary Yellowstone caldera and surrounding areas are world famous for their spectacular displays of hydrothermal activity, i.e., geysers, fumeroles, and hot springs that suggest shallow sources of heat. Low seismic velocities, high attenuation, and shallow earthquake loci corroborate an anomalous crustal structure associated with the Yellowstone caldera. Combined with the three-dimensional seismic velocity structure, the inverted three-dimensional Q structure discussed in this paper provides additional information on the unusual properties of this active silicic volcanic region. The Yellowstone caldera (noted by III in Figure 1), measuring 45 km by 70 km, was formed 600,000 years ago by the catastrophic eruption of 1000 km 3 of rhyolitic ash

We discuss the maximum entropy approach to obtaining the weights associated with the ordered weighted averaging (OWA) aggregation operator. The resulting weights are called the MEOWA weights. Using the method of LaGrange multipliers, we obtain an analytic form for these weights and describe some of their properties. The concept of immediate probabilities is introduced as being a transformation of a probability distribution based on a decision maker's degree of optimism. This transformation, which is affected by an OWA ...

A method for measuring mooring line tension is proposed based on observation of the natural frequencies of the mooring line segment between the winch and the fairlead. The anchor line tension is observed through the string equation where an analytical expression for the line's eigenfrequencies is obtained. The tension is observed on line by utilizing a nonparametric system identification approach in which the peaks of a maximum entropy spectrum of the transverse acceleration measures of the vibrating string are automatically identified. The method is verified against full-scale data from the Troll B floating concrete oil production platform operating in the North Sea.

This paper describes an analytical sensitivity analysis and optimization implementation for cable-stayed bridge design. The finite element software is based on the Vax/VMS version of the Modulef code [I MODULEF Reference Guide. INRIA (1992).] and was adapted to an IBM-PC compatible. The main focus of this research concerns the analytical sensitivity analysis developed on this platform. The cable-stayed bridge optimization is posed as a multiobjective optimization with goals of minimum cost of material, stresses and displacements. Cable anchor positions on the main girder and pylon and cross-sectional sizes of the structural members are dealt with as design variables. By using the maximum entropy formalism it is shown that a Pareto solution may be found indirectly by the unconstrained optimization of a scalar function. The validity and effectiveness of the proposed technique is examined by means of a three-span steel cable-stayed bridge. 0 1997 Civil-

Effective human and automatic processing of speech requires recovery of more than just the words. It also involves recovering phenomena such as sentence boundaries, filler words, and disfluencies, referred to as structural metadata. We describe a metadata detection system that combines information from different types of textual knowledge sources with information from a prosodic classifier. We investigate maximum entropy and conditional random field models, as well as the predominant HMM approach, and find that discriminative models generally outperform generative models. We report system performance on both broadcast news and conversational telephone speech tasks, illustrating significant performance differences across tasks and as a function of recognizer performance. The results represent the state of the art, as assessed in the NIST RT-04F evaluation.

Gradual development of anomalies (i.e., deviations from the nominal condition) may alter the quasi-static behavior patterns of human-engineered complex systems. This chapter presents a recently reported technique of pattern recognition, called Symbolic Dynamic Filtering (SDF), for early detection and prognosis of such changes in behavior patterns due to slowly evolving anomalies that may be benign or malignant. The underlying concept of SDF is built upon the principles of Symbolic Dynamics, Information Theory, and Statistical Signal Processing, where time series data from selected sensor(s) in the fast time scale of the process dynamics are analyzed at discrete epochs in the slow time scale of anomaly evolution. The key idea here is early detection and identification of (possible) changes in quasi-static statistical patterns of the dynamical system behavior. An important feature of this pattern recognition technique is extraction of the relevant statistics by conversion of the time series data into a symbol sequence by appropriate coarse-graining of the imbedded information. As an alternative to the currently practiced method of phase-space partitioning in the time domain, a new concept of partitioning is introduced for symbol generation, based on wavelet analysis of the time series data. This chapter also discusses various aspects of the wavelet-based partitioning tool, such as selection of the wavelet basis, noise mitigation, and robustness to spurious disturbances. The partitioning scheme is built upon the principle of maximum entropy such that the regions of the data space with more information are partitioned finer and those with sparse information are partitioned coarser. The algorithms of SDF are constructed to solve two problems: (i) Forward problem of Pattern Recognition for (offline) characterization of the anomalous behavior, relative to the nominal behavior; and (ii) Inverse problem of Pattern * Identification for (online) estimation of parametric or non-parametric changes based on the knowledge assimilated in the forward problem and the observed time series data of quasi-stationary process response.

Effective Management and conservation of wildlife populations and their habitats largely depend on our ability to understand and predict species-habitat interactions. Intensive ground surveys cannot keep pace with the rate of land-use change and consequently habitat composition over large areas. We explored how effectively do Remote sensing satellite imagery and GIS modeling technique could be used for assessing habitat suitability of Rhinoceros unicornis and what are the habitat factors influencing rhinoceros distribution in lowland floodplain of Nepal. The Landsat ETM+ satelli te imagery (Path-142& Row-041) of the study area was used for classifying land use/land cover. Image processing and feature extraction was done in Erdas Imagine 8.7. We used supervised classification technique with 350 training points and 40 test points. GI S layers of habitat parameters-continuous distances from grasslands, water body, guard posts, Agriculture/ settlement and categorical land use/land cover map were used as predictor variables. The points of animal presence location were used as suitable pr oxies and Maximum entropy (MAXENT) modeling was run for predicting species potential geographic distribution. The most significant result of the image classification was that the proportion of pure grassland patches in the chitwan National Park is only 7 percent of the total area. Riparian forest, developed as a result of grassland succession, accounts for 8 percent of the park area which otherwise, as a grassland, served as a food source for R. unicornis. The Maxent model based on remotely sensed factors, habitat factors and rhino presence locations resulted in much larger area classified as suitable for Rhinos. The contribution of the variable distance to water (48.6 percent) was highest to impact the model. The model performance was assessed using receiver operating Characteristics (ROC) plots and Jackknife tests.

Recent research has shown the benefit of framing problems of imitation learning as solutions to Markov Decision Problems. This approach reduces learning to the problem of recovering a utility function that makes the behavior induced by a near-optimal policy closely mimic demonstrated behavior. In this work, we develop a probabilistic approach based on the principle of maximum entropy. Our approach provides a well-defined, globally normalized distribution over decision sequences, while providing the same performance guarantees as existing methods. We develop our technique in the context of modeling realworld navigation and driving behaviors where collected data is inherently noisy and imperfect. Our probabilistic approach enables modeling of route preferences as well as a powerful new approach to inferring destinations and routes based on partial trajectories.

When Shannon entropy is used as a criterion in the optimal design of experiments, advantage can be taken of the classical identity representing the joint entropy of parameters and observations as the sum of the marginal entropy of the observations and the preposterior conditional entropy of the parameters. Following previous work in which this idea was used in spatial sampling, the method is applied to standard parameterized Bayesian optimal experimental design. Under suitable conditions, which include non-linear as well as linear regression models, it is shown in a few steps that maximizing the marginal entropy of the sample is equivalent to minimizing the preposterior entropy, the usual Bayesian criterion, thus avoiding the use of conditional distributions. It is shown using this marginal formulation that under normality assumptions every standard model which has a two-point prior distribution on the parameters gives an optimal design supported on a single point. Other results include a new asymptotic formula which applies as the error variance is large and bounds on support size.

Traditionally, the MaxEnt workshops start by a tutorial day. This paper summarizes my talk during 2001'th workshop at John Hopkins University. The main idea in this talk is to show how the Bayesian inference can naturally give us all the necessary tools we need to solve real inverse problems: starting by simple inversion where we assume to know exactly the forward model and all the input model parameters up to more realistic advanced problems of myopic or blind inversion where we may be uncertain about the forward model and we may have noisy data.

The availability of detailed environmental data, together with inexpensive and powerful computers, has fueled a rapid increase in predictive modeling of species environmental requirements and geographic distributions. For some species, detailed presence/absence occurrence data are available, allowing the use of a variety of standard statistical techniques. However, absence data are not available for most species. In this paper, we introduce the use of the maximum entropy method (Maxent) for modeling species geographic distributions with presence-only data. Maxent is a general-purpose machine learning method with a simple and precise mathematical formulation, and it has a number of aspects that make it well-suited for species distribution modeling. In order to investigate the efficacy of the method, here we perform a continental-scale case study using two Neotropical mammals: a lowland species of sloth, Bradypus variegatus, and a small montane murid rodent, Microryzomys minutus. We compared Maxent predictions with those of a commonly used presence-only modeling method, the Genetic Algorithm for Rule-Set Prediction (GARP). We made predictions on 10 random subsets of the occurrence records for both species, and then used the remaining localities for testing. Both algorithms provided reasonable estimates of the species' range, far superior to the shaded outline maps available in field guides. All models were significantly better than random in both binomial tests of omission and receiver operating characteristic (ROC) analyses. The area under the ROC curve (AUC) was almost always higher for Maxent, indicating better discrimination of suitable versus unsuitable areas for the species. The Maxent modeling approach can be used in its present form for many applications with presence-only datasets, and merits further research and development.

Through the growth of common networking aera and its development, Internet has developed a capable stage for connected knowledge, replacing concepts and distribution sentiments. Common media covers an enormous quantity of the sentimentality information in the procedure of twitters, blogs, and informs on the position, posts, etc. In this paper, the maximum general micro blogging stage Twitter is used. Twitter sentiment study is a request of sentimentality scrutiny on information since Twitter (tweets), to excerpt user’s sentiments and opinions. The key goal is to discover in what way script study methods can be used to crack into approximately of the information in a sequence of markers concentrating on changed movements of tweets dialects, tweets dimensions on twitter. New valuations display that the future machine learning classifiers are effective and do improved in relations of correctness and period. The future algorithm is executed in python

Racehorses given a slurry of sodium bicarbonate (known colloquially as a 'milkshake') before a race have elevated concentrations of carbon dioxide in their blood. Racing administrators have reacted to this attempt to enhance the performance of the animal by setting limits to 'total carbon dioxide' (TCO2, the sum of carbon dioxide, carbonic acid, carbonate and bicarbonate) in prerace samples of plasma. The threshold limit for TCO2 in the rules of racing is an amount concentration of 36.0 mmol/L, with further action ensuing if the reported concentration is greater than an action limit that is calculated from the knowledge of the measurement uncertainty. At present in Australia, the action limit is 37.0 mmol/L, which is based on a combined standard measurement uncertainty of 0.22 mmol/L. From data obtained in a 1997 study in 515 normal racehorses, we have established the distribution (as a probability density function, PDF) of TCO2. This is combined with data from Australian laboratories of 126 horses that were tested following a positive screen, out of which 78 were confirmed positive. We employ the maximum entropy method to establish the PDFs and then apply Bayes Theorem to answer the question 'given the measured TCO2 concentration what is the probability that a horse has been administered bicarbonate'? The distributions are not normal, which precludes simple approaches that calculate standard deviations from the data. For an action level of 37.0 mmol/L, there is a chance of only 1 in 2 020 000 that a nondoped horse will be judged to be doped, which implies this present threshold is unlikely to lead to conviction of an innocent trainer.

The 2005 Conference on Computational Natural Language Learning (CoNLL-2005) is the ninth in a series of meetings organized by SIGNLL, the ACL special interest group on natural language learning. This year's CoNLL will be held in Ann Arbor, Michigan, on June 29 and 30, in conjunction with the ACL 2005 conference.

The two component extreme value (TCEV) distribution has recently been shown to account for most of the characteristics of the real flood experience. A new method of parameter estimation for this distribution is derived using the principle of maximum entropy (POME). This method of parameter estimation is suitable for application in both the site-specific and regional cases and appears simpler than the maximum likelihood estimation method. Statistical properties of the regionalized estimation were evaluated using a Monte Carlo approach and compared with those of the maximum likelihood regional estimators.

This paper develops a methodology for quantifying model risk in quantile risk estimates. The application of quantile estimates to risk assessment has become common practice in many disciplines, including hydrology, climate change, statistical process control, insurance and actuarial science and the uncertainty surrounding these estimates has long been recognized. Our work is particularly important in finance, where quantile estimates (called Value-at-Risk) have been the cornerstone of banking risk management since the mid 1980's. A recent amendment to the Basel II Accord recommends additional market risk capital to cover all sources of 'model risk' in the estimation of these quantiles. We provide a novel and elegant framework whereby quantile estimates are adjusted for model risk, relative to a benchmark which represents the state of knowledge of the authority that is responsible for model risk. A simulation experiment in which the degree of model risk is controlled illustrates how to quantify model risk and compute the required regulatory capital add-on for banks. An empirical example based on real data shows how the methodology can be put into practice, using only two time series (daily Value-at-Risk and daily profit and loss) from a large bank. We conclude with a discussion of potential applications to non-financial risks.

This paper proposes a solution to the spectral color constancy problem. The method is based on a statistical model for the surface reflectance spectrum and applies a maximum entropy constraint. Unlike prior methods based on linear models, the solution process does not require a set of basis functions to be defined, nor does it require a database of spectra to be specified in advance. Experiments on simulated and real data show that spectral estimation using the maximum entropy approach is feasible and performs similarly to existing spectral methods in spite of the lower level of a priori information required.

Absrmct-An algorithm for the solution of the h e a r equations for the "covariance method" of linear prediction is stated and proved. The algorithm requires only 0 (p2) arithmetic operations, and in form resembles the Levinson algorithm for solution of the linear equations for the "correlation method" of linear prediction.

A comparative study of the magnetocaloric effect (MCE) is reported in two different types of chemically synthesized magnetic nanoparticle systems-cobalt ferrite and manganese zinc ferrite with mean size around 5 and 15 nm, respectively. While CoFe 2 O 4 nanoparticles were synthesized using co-precipitation, the Mn 0.68 Zn 0.25 Fe 2.07 O 4 (MZFO) nanoparticles were prepared by reverse micelle technique using AOT as surfactant. Our results indicate that the change in entropy with the change in applied magnetic field (dS/dH) is reasonably large for this class of nanoparticles and has a wide distribution over a broad temperature range covering the region above and below the blocking temperature. The maximum entropy change is influenced by the particle size, overall distribution in anisotropy and magnetic moments. r

The biodiversity scaling metrics widely studied in macroecology include the species-area relationship (SAR), the scale-dependent species-abundance distribution (SAD), the distribution of masses or metabolic energies of individuals within and across species, the abundance-energy or abundance-mass relationship across species, and the species-level occupancy distributions across space. We propose a theoretical framework for predicting the scaling forms of these and other metrics based on the state-variable concept and an analytical method derived from information theory. In statistical physics, a method of inference based on information entropy results in a complete macro-scale description of classical thermodynamic systems in terms of the state variables volume, temperature, and number of molecules. In analogy, we take the state variables of an ecosystem to be its total area, the total number of species within any specified taxonomic group in that area, the total number of individuals across those species, and the summed metabolic energy rate for all those individuals. In terms solely of ratios of those state variables, and without invoking any specific ecological mechanisms, we show that realistic functional forms for the macroecological metrics listed above are inferred based on information entropy. The Fisher log series SAD emerges naturally from the theory. The SAR is predicted to have negative curvature on a log-log plot, but as the ratio of the number of species to the number of individuals decreases, the SAR becomes better and better approximated by a power law, with the predicted slope z in the range of 0.14-0.20. Using the 3/4 power mass-metabolism scaling relation to relate energy requirements and measured body sizes, the Damuth scaling rule relating mass and abundance is also predicted by the theory. We argue that the predicted forms of the macroecological metrics are in reasonable agreement with the patterns observed from plant census data across habitats and spatial scales. While this is encouraging, given the absence of adjustable fitting parameters in the theory, we further argue that even small discrepancies between data and predictions can help identify ecological mechanisms that influence macroecological patterns.

We consider the problem of classifying documents not by topic, but by overall sentiment, e.g., determining whether a review is positive or negative. Using movie reviews as data, we find that standard machine learning techniques definitively outperform human-produced baselines. However, the three machine learning methods we employed (Naive Bayes, maximum entropy classification, and support vector machines) do not perform as well on sentiment classification as on traditional topic-based categorization. We conclude by examining factors that make the sentiment classification problem more challenging. 1

The "covariance" of complex random variables and processes, when defined consistently with the corresponding notion for real random variables, is shown to be detc!rmioed by the usual (complex) covariance together with a quantity called the pseudo-covariance. A characterization of uncorrelatedness and wide-sense stationarity in terms of covariance and pseudocovariance is given. Complex random variables and processes with a vanishing pseudo-covariaqce are called proper. It is shown that properness is preserved under affine transformations and that the complex-multivariate Gaussian density assumes a natural form only for proper random variables. The maximumsotropy theorem is generalized to the complex-multivariate case. The differential entropy of a complex random vector with a tixed correlation matrix is shown to be maximum, if and only if the random vector is proper, Gaussian and zero-mean. The notion of chulur si'utio-is introduced. For the class of proper complex random processes, a discrete Fourier transform correspondence is derived dating circular stationarity in the time domain to uncomlatedness in the fmquency domain. As an application of the theory, the capacity of a discrete-time channel with complex inputs, proper complex additive white Gaussian noise, and a finite complex unit-sample response is determined. This derivation is considerably simpler than an earlier derivation for the red discrete-time Gaussian channel with intersymbol interference, whose capacity is obtained as a by-product of the results for the complex channel.

The NMRPipe system is a UNIX software environment of processing, graphics, and analysis tools designed to meet current routine and research-oriented multidimensional processing requirements, and to anticipate and accommodate future demands and developments. The system is based on UNIX pipes, which allow programs running simultaneously to exchange streams of data under user control. In an NMRPipe processing scheme, a stream of spectral data flows through a pipeline of processing programs, each of which performs one component of the overall scheme, such as Fourier transformation or linear prediction. Complete multidimensional processing schemes are constructed as simple UNIX shell scripts. The processing modules themselves maintain and exploit accurate records of data sizes, detection modes, and calibration information in all dimensions, so that schemes can be constructed without the need to explicitly define or anticipate data sizes or storage details of real and imaginary channels during processing. The asynchronous pipeline scheme provides other substantial advantages, including high flexibility, favorable processing speeds, choice of both all-in-memory and disk-bound processing, easy adaptation to different data formats, simpler software development and maintenance, and the ability to distribute processing tasks on multi-CPU computers and computer networks. nmrPipe -fn MEM -ndim 2 -neg -zero -alpha 0.001 -xconv EM -xcQl 20 -yconv EM -ycQl 15 -sigma 200 -freq pipe2xyz -out ft/noe%03d.ft3 -z -inPlace \ Read Vectors from X-Axis \ Solvent Filter \

Hydrogen-enriched methane flame Second law analysis Propagating triple flame a b s t r a c t A theoretical-numerical analysis based on the second law of thermodynamics is used to examine the propagation of laminar H 2 -enriched CH 4 -air flames. The analysis is based on computing the various entropy generation terms in a transient reacting flow field. A comprehensive, time-dependent computational model, which employs a detailed description of chemistry and transport, is used to simulate the transient ignition and flame propagation in this reacting flow field. Flames are ignited in a jet-mixing layer far downstream of the burner. Following ignition, a well-defined triple flame is formed that propagates upstream with nearly constant flame displacement speed along the stoichiometric mixture fraction line. As the flame approaches the burner, it transitions to a double flame, and subsequently to a burner-stabilized nonpremixed flame. The triple point exhibits the maximum entropy generation, indicating that this point is characterized by high chemical reactivity, as well as large temperature and mass fraction gradients. The volumetric

In this paper, we describe a novel uncertaintybased technique for predicting the future motions of a moving person. Our model assumes that people behave purposefully -efficiently acting to reach intended destinations. We employ the Markov decision process framework and the principle of maximum entropy to obtain a probabilistic, approximately optimal model of human behavior that admits efficient inference and learning algorithms. The method learns a cost function of features of the environment that explains previously observed behavior. This enables generalization to physical changes in the environment, and entirely different environments. Our approach enables robots to plan paths that balance time-togoal and pedestrian disruption. We quantitatively show the improvement provided by our approach.

In this paper we introduce the class Weibull power series (WPS) of distributions which is obtained by compounding Weibull and power series distributions, where compounding procedure follows same way that was previously carried out by . This new class of distributions has as particular case the two-parameter class exponential power series (EPS) of distributions, which was introduced recently by Chahkandi and Ganjali (2009). Like EPS distributions, our class contains several distributions which have been introduced and studied such as: exponential geometric (Adamidis and Loukas, 1998), exponential Poisson and exponential logarithmic (Tahmasbi and Rezaei, 2008) distributions. Moreover, the hazard function of our class may take increasing, decreasing, upside down bathtub forms, among others, while hazard function of a EPS distribution is only decreasing. We obtain several properties of the WPS distributions such as moments, order statistics, estimation by maximum likelihood and inference for large sample. Furthermore, EM algorithm is also used to determine the maximum likelihood estimates of the parameters and we discuss maximum entropy characterizations under suitable constraints. Special distributions are studied in some details. Applications to real data sets are given to show the flexibility and potentiality of the proposed class of distributions.

Pan troglodytes verus is considered one of the most endangered primates in the world due to habitat loss or degradation, and it is considered extinct in several regions of West-Africa. This study aims at developing a model that predicts habitat suitability for chimpanzee under different land cover scenarios and supports a multi-temporal analysis of the recent habitat evolution in southern Guinea-Bissau. The model, built using a maximum entropy framework, is spatially explicit, uses eco-geographic variables derived from landscape compositional, structural and functional attributes, and is based on nest location data collected in the field. The results show that the developed model has good levels of reliability and the map produced through its application reveals a current extension of 752 km 2 of adequate habitat in a 2723 km 2 study area. Forest related EGVs (ecogeographical variables) had the higher importance for the model production. Habitat suitability maps produced for three dates in the last three decades show that there has been a marked decrease in habitat extension and in connectivity to neighbouring regions. We analyse the implications of these results in possible strategies for the conservation of chimpanzees in Guinea-Bissau.

In the construction of a part-of-speech annotated corpus, we are constrained by a fixed budget. A fully annotated corpus is required, but we can afford to label only a subset. We train a Maximum Entropy Markov Model tagger from a labeled subset and automatically tag the remainder. This paper addresses the question of where to focus our manual tagging efforts in order to deliver an annotation of highest quality. In this context, we find that active learning is always helpful. We focus on Query by Uncertainty (QBU) and Query by ...

We discuss a Statistical Mechanics approach in the manner of Edwards to the "inherent states" (defined as the stable configurations in the potential energy landscape) of glassy systems and granular materials. We show that at stationarity the inherent states are distributed according a generalized Gibbs measure obtained assuming the validity of the principle of maximum entropy, under suitable constraints. In particular we consider three lattice models (a diluted Spin Glass, a monodisperse hard-sphere system under gravity and a hard-sphere binary mixture under gravity) undergoing a schematic "tap dynamics", showing via Monte Carlo calculations that the time average of macroscopic quantities over the tap dynamics and over such a generalized distribution coincide. We also discuss about the general validity of this approach to non thermal systems.

Random numbers are used in a wide variety of applications from simulation and encryption to gambling and clinical trials. A good quality random number generator is an asset for applications like encryption, randomized designs and network and information security. Various mathematical models have been developed in the past to improve the quality of random numbers. It can be construed that in general to obtain random numbers of excellent quality, a complex mathematical model has to be used which can be a performance bottleneck. In this work, we propose a novel technique to implement a True Random Number Generator (TRNG) using sources of uncertainty found within Graphics Processing Units (GPUs) together with histogram equalization to obtain maximum entropy. We evaluate the random numbers generated by our approach using four tests. First, we measure the correlation values between two sequences of random numbers, second, we measure the entropy values, third, we use watermarking, an application used in network security and finally we use Monte Carlo analysis for pi-value calculation. Based on these quality measurements, our method has achieved better results than popular random number generators compared in this work. Furthermore, this approach is a massively scalable solution ideal for high performance computing implementations.

I explore the possibility that the laws of physics might be laws of inference rather than laws of nature. What sort of dynamics can one derive from well-established rules of inference? Specifically, I ask: Given relevant information codified in the initial and the final states, what trajectory is the system expected to follow? The answer follows from a principle of inference, the principle of maximum entropy, and not from a principle of physics. The entropic dynamics derived this way exhibits some remarkable formal similarities with other generally covariant theories such as general relativity.

,-The application of the maximum irreversible entropy variation principle to the calculation of the lattice constant is suggested as a new methodological approach, PACS 61.90-Other topics in structure of solids and liquids, PACS 63,70-Statistical mechanics of lattice vibrations and displacive phase transitions, PACS 63,90-Other topics in lattice dynamiCS, PACS 05,90-Other topics in statistical physics and thermodynamics,

Topographic map algorithms that are aimed at building "faithful representations" also yield maps that transfer the maximum amount of information available about the distribution from which they receive input. The weight density (magnification factor) of these maps is proportional to the input density, or the neurons of these maps have an equal probability to be active (equiprobabilistic map). As MSE minimization is not compatible with equiprobabilistic map formation in general, a number of heuristics have been devised in order to compensate for this discrepancy in competitive learning schemes, e.g. by adding a "conscience" to the neurons' firing behavior. However, rather than minimizing a modified MSE criterion, we introduce a new unsupervised competitive learning rule, called the kernel-based Maximum Entropy learning Rule (kMER), for topographic map formation, that optimizes an information-theoretic criterion directly. To each neuron a radially symmetric kernel is associated, with a given center and radius, and the two are updated in such a way that the (unconditional) information-theoretic entropy of the neurons' outputs is maximized. We review a number of competitive learning rules for building equiprobabilistic maps. As benchmark tests for the faithfulness of the representations, we consider two types of distributions and compare the performances of these rules and kMER, for batch and incremental learning. As a first example application, we consider non-parametric density estimation where the maps are used for generating "pilot" estimates in kernel-based density estimation. The second application we envisage for kMER is "on-line" adaptive filtering of speech signals, using Gabor functions as wavelet filters. The topographic feature maps that are developed in this way differ in several respects from those obtained with Kohonen's Adaptive-Subspace SOM algorithm.