Downloads & Free Reading Options - Results

In this paper we borrow concepts from Information Theory and Statistical Mechanics to perform a pattern recognition procedure on a set of x-ray hazelnut images. We identify two relevant statistical scales, whose ratio affects the performance of a machine learning algorithm based on statistical observables, and discuss the dependence of such scales on the image resolution. Finally, by averaging the performance of a Support Vector Machines algorithm over a set of training samples, we numerically verify the predicted onset of an optimal scale of resolution, at which the pattern recognition is favoured.

In this paper we borrow concepts from Information Theory and Statistical Mechanics to perform a pattern recognition procedure on a set of x-ray hazelnut images. We identify two relevant statistical scales, whose ratio affects the performance of a machine learning algorithm based on statistical observables, and discuss the dependence of such scales on the image resolution. Finally, by averaging the performance of a Support Vector Machines algorithm over a set of training samples, we numerically verify the predicted onset of an optimal scale of resolution, at which the pattern recognition is favoured.

In this paper we borrow concepts from Information Theory and Statistical Mechanics to perform a pattern recognition procedure on a set of x-ray hazelnut images. We identify two relevant statistical scales, whose ratio affects the performance of a machine learning algorithm based on statistical observables, and discuss the dependence of such scales on the image resolution. Finally, by averaging the performance of a Support Vector Machines algorithm over a set of training samples, we numerically verify the predicted onset of an optimal scale of resolution, at which the pattern recognition is favoured.

In this paper we borrow concepts from Information Theory and Statistical Mechanics to perform a pattern recognition procedure on a set of x-ray hazelnut images. We identify two relevant statistical scales, whose ratio affects the performance of a machine learning algorithm based on statistical observables, and discuss the dependence of such scales on the image resolution. Finally, by averaging the performance of a Support Vector Machines algorithm over a set of training samples, we numerically verify the predicted onset of an optimal scale of resolution, at which the pattern recognition is favoured.

Further upgrades have been made to a previously developed statistical pattern recognition and analysis software tool. This tool, referred to as STATPACK, has been developed in the MATLAB processing environment. The upgrades include Fisher discriminant projection capabilities, for data structure analysis.

The development of a MATLAB based statistical pattern recognition software package (referred to as STATPACK) was begun. Initial developments include a two-dimensional coordinate projection capability, and a two-dimensional eigenvector projection capability. The operating system's directory structure has been utilized to allow for multiple levels of data separation algorithms in subsequent development.

This paper examines the current status of the statistical pattern recognition by the topics: classification rules, feature extraction, contextual analysis, etc. Important but unsolved problem areas are also explored. The relationship between the statistical pattern recognition and signal processing is also considered.

Probabilistic learning strategies currently use are inefficient, requiring high classifier complexity and large training samples. In this report, we introduce and analyze an asymptotically efficient differential learning strategy. It guarantees the best generalization allowed by the chosen classifier paradigm. Differential learning also requires the classifier with minimal complexity. The theory is demonstrated in several real-world machine learning/pattern recognition tasks. Learning, Pattern recognition, Classification, Neural networks.

Probabilistic learning strategies currently use are inefficient, requiring high classifier complexity and large training samples. In this report, we introduce and analyze an asymptotically efficient differential learning strategy. It guarantees the best generalization allowed by the chosen classifier paradigm. Differential learning also requires the classifier with minimal complexity. The theory is demonstrated in several real-world machine learning/pattern recognition tasks. Learning, Pattern recognition, Classification, Neural networks.

A necessary condition for a real valued Frechet differentiable function of a vector variable have an extremum at a vector x sub 0 is that the Frechet derivative vanishes at x sub 0. A relationship between Frechet differentials and matrix derivatives was established that obtains a necessary condition on the matrix derivative at an extrema. These results are applied to various scalar functions of matrix variables which occur in statistical pattern recognition.

A necessary condition for a real valued Frechet differentiable function of a vector variable have an extremum at a vector x sub 0 is that the Frechet derivative vanishes at x sub 0. A relationship between Frechet differentials and matrix derivatives was established that obtains a necessary condition on the matrix derivative at an extrema. These results are applied to various scalar functions of matrix variables which occur in statistical pattern recognition.

A necessary condition for a real valued Frechet differentiable function of a vector variable have an extremum at a vector x sub 0 is that the Frechet derivative vanishes at x sub 0. A relationship between Frechet differentials and matrix derivatives was established that obtains a necessary condition on the matrix derivative at an extrema. These results are applied to various scalar functions of matrix variables which occur in statistical pattern recognition.

A necessary condition for a real valued Frechet differentiable function of a vector variable have an extremum at a vector x sub 0 is that the Frechet derivative vanishes at x sub 0. A relationship between Frechet differentials and matrix derivatives was established that obtains a necessary condition on the matrix derivative at an extrema. These results are applied to various scalar functions of matrix variables which occur in statistical pattern recognition.

The false discovery rate (FDR) and false nondiscovery rate (FNDR) have received considerable attention in the literature on multiple testing. These performance measures are also appropriate for classification, and in this work we develop generalization error analyses for FDR and FNDR when learning a classifier from labeled training data. Unlike more conventional classification performance measures, the empirical FDR and FNDR are not binomial random variables but rather a ratio of binomials, which introduces challenges not addressed in conventional analyses. We develop distribution-free uniform deviation bounds and apply these to obtain finite sample bounds and strong universal consistency.

This report presents a summary of the basic principles of pattern recognition and statistical decision theory and applies them to the problem of classifying radar returns. While pattern recognition techniques have been applied to radar signal detection problems, they have rarely been used in testing hypothesis for classifying radar returns. Two techniques, the parametric Bayes and the non-parametric K-Nearest Neighbor algorithms, were compared using simulated radar backscatter data. The error rate of these algorithms was the chief criterion used for the evaluation of performance. The results showed that the Nearest Neighbor technique gives a smaller error rate than the Bayes technique for the limited data sets tested. (Author)

This report presents a summary of the basic principles of pattern recognition and statistical decision theory and applies them to the problem of classifying radar returns. While pattern recognition techniques have been applied to radar signal detection problems, they have rarely been used in testing hypothesis for classifying radar returns. Two techniques, the parametric Bayes and the non-parametric K-Nearest Neighbor algorithms, were compared using simulated radar backscatter data. The error rate of these algorithms was the chief criterion used for the evaluation of performance. The results showed that the Nearest Neighbor technique gives a smaller error rate than the Bayes technique for the limited data sets tested. (Author)

Driving styles have a great influence on vehicle fuel economy, active safety, and drivability. To recognize driving styles of path-tracking behaviors for different divers, a statistical pattern-recognition method is developed to deal with the uncertainty of driving styles or characteristics based on probability density estimation. First, to describe driver path-tracking styles, vehicle speed and throttle opening are selected as the discriminative parameters, and a conditional kernel density function of vehicle speed and throttle opening is built, respectively, to describe the uncertainty and probability of two representative driving styles, e.g., aggressive and normal. Meanwhile, a posterior probability of each element in feature vector is obtained using full Bayesian theory. Second, a Euclidean distance method is involved to decide to which class the driver should be subject instead of calculating the complex covariance between every two elements of feature vectors. By comparing the Euclidean distance between every elements in feature vector, driving styles are classified into seven levels ranging from low normal to high aggressive. Subsequently, to show benefits of the proposed pattern-recognition method, a cross-validated method is used, compared with a fuzzy logic-based pattern-recognition method. The experiment results show that the proposed statistical pattern-recognition method for driving styles based on kernel density estimation is more efficient and stable than the fuzzy logic-based method.

A review along with criticism of some recent work on nonparametric and sequential rules in statistical pattern recognition is given in this paper. Some new results and directions for future work are also discussed. (Author)

A review along with criticism of some recent work on nonparametric and sequential rules in statistical pattern recognition is given in this paper. Some new results and directions for future work are also discussed. (Author)

A review along with criticism of some recent work on nonparametric and sequential rules in statistical pattern recognition is given in this paper. Some new results and directions for future work are also discussed. (Author)

This article is from International Journal of Oncology , volume 45 . Abstract To predict precisely severe toxicity of irinotecan, we evaluated the association of UGT1A variants, haplotypes and the combination of UGT1A genotypes to severe toxicity of irinotecan. UGT1A1*6 (211G>A), UGT1A1*28 (TA6>TA7), UGT1A1*60 (−3279T>G), UGT1A7 (387T>G), UGT1A7 (622T>C), and UGT1A9*1b (−118T9>T10, also named *22) were genotyped in 123 patients with metastatic colorectal cancer who had received irinotecan-based chemotherapy. Among the 123 patients, 73 were enrolled in either of two phase II studies of the FOLFIRI (leucovorin, 5-fluorouracil and irinotecan) regimen; these patients constituted the training population, which was used to construct the predicting system. The other 50 patients constituted the validation population; these 50 patients either had participated in a phase II study of irinotecan/5′-deoxy-5-fluorouridine or were among consecutive patients who received FOLFIRI therapy. This prediction system used sequential forward floating selection based on statistical pattern recognition using UGT1A genotypes, gender and age. Several UGT1A genotypes [UGT1A1*6, UGT1A7 (387T>G), UGT1A7 (622T>C) and UGT1A9*1b] were associated with the irinotecan toxicity. Among the haplotypes, haplotype-I (UGT1A1: −3279T, TA6, 211G; UGT1A7: 387T, 622T; UGT1A9: T10) and haplotype-II (UGT1A1: −3279T, TA6, 211A; UGT1A7: 387G, 622C; UGT1A9: T9) were also associated with irinotecan toxicity. Furthermore, our new system for predicting the risk of irinotecan toxicity was 83.9% accurate with the training population and 72.1% accurate with the validation population. Our novel prediction system using statistical pattern recognition depend on genotypes in UGT1A, age and gender; moreover, it showed high predictive performance even though the treatment regimens differed among the training and validation patients.

Acoustic emission (AE) data were acquired during fatigue testing of an aluminum 2024-T4 compact tension specimen using a commercially available AE system. AE signals from crack extension were identified and separated from noise spikes, signals that reflected from the specimen edges, and signals that saturated the instrumentation. A commercially available software package was used to train a statistical pattern recognition system to classify the signals. The software trained a network to recognize signals with a 91-percent accuracy when compared with the researcher's interpretation of the data. Reasons for the discrepancies are examined and it is postulated that additional preprocessing of the AE data to focus on the extensional wave mode and eliminate other effects before training the pattern recognition system will result in increased accuracy.

Traditional natural language parsers are based on rewrite rule systems developed in an arduous, time-consuming manner by grammarians. A majority of the grammarian's efforts are devoted to the disambiguation process, first hypothesizing rules which dictate constituent categories and relationships among words in ambiguous sentences, and then seeking exceptions and corrections to these rules. In this work, I propose an automatic method for acquiring a statistical parser from a set of parsed sentences which takes advantage of some initial linguistic input, but avoids the pitfalls of the iterative and seemingly endless grammar development process. Based on distributionally-derived and linguistically-based features of language, this parser acquires a set of statistical decision trees which assign a probability distribution on the space of parse trees given the input sentence. These decision trees take advantage of significant amount of contextual information, potentially including all of the lexical information in the sentence, to produce highly accurate statistical models of the disambiguation process. By basing the disambiguation criteria selection on entropy reduction rather than human intuition, this parser development method is able to consider more sentences than a human grammarian can when making individual disambiguation rules. In experiments between a parser, acquired using this statistical framework, and a grammarian's rule-based parser, developed over a ten-year period, both using the same training material and test sentences, the decision tree parser significantly outperformed the grammar-based parser on the accuracy measure which the grammarian was trying to maximize, achieving an accuracy of 78% compared to the grammar-based parser's 69%.

Traditional natural language parsers are based on rewrite rule systems developed in an arduous, time-consuming manner by grammarians. A majority of the grammarian's efforts are devoted to the disambiguation process, first hypothesizing rules which dictate constituent categories and relationships among words in ambiguous sentences, and then seeking exceptions and corrections to these rules. In this work, I propose an automatic method for acquiring a statistical parser from a set of parsed sentences which takes advantage of some initial linguistic input, but avoids the pitfalls of the iterative and seemingly endless grammar development process. Based on distributionally-derived and linguistically-based features of language, this parser acquires a set of statistical decision trees which assign a probability distribution on the space of parse trees given the input sentence. These decision trees take advantage of significant amount of contextual information, potentially including all of the lexical information in the sentence, to produce highly accurate statistical models of the disambiguation process. By basing the disambiguation criteria selection on entropy reduction rather than human intuition, this parser development method is able to consider more sentences than a human grammarian can when making individual disambiguation rules. In experiments between a parser, acquired using this statistical framework, and a grammarian's rule-based parser, developed over a ten-year period, both using the same training material and test sentences, the decision tree parser significantly outperformed the grammar-based parser on the accuracy measure which the grammarian was trying to maximize, achieving an accuracy of 78% compared to the grammar-based parser's 69%.

It has been widely known that data fusion, neural network and statistical pattern recognition technologies can be applied to target identification and classification problems. The main objective of this paper is to find out which of these techniques would be easy to use and provide acceptable results. We had selected the Multi-sensor Correlation Model 1 from the field of data fusion technology. The concept of this model is based on the coefficient of similarity. For target identification problem, one have to estimate the coefficient of similarity between a known target (X) and the target (Y) to be identified. If the coefficient is closed to one , then it implied that target (Y) is the same as target (X), otherwise if the coefficient is close to zero, then it implied that target (Y) is not the same as target (X). It is mathematical simple and easy to implement. The Bayesian Model 2 was selected from the field of statistical pattern recognition technology, This is a conditional probability model. For target identification problem, one have to calculate the posterior probability of a known target (X) given the target (Y) to one to be identified. If the conditional probability is close to one , then it implied that target (x) and target (Y) is the same, otherwise if it is close to zero, then it implied that target(X) and target(Y) is not the same. This model required multivariate normal assumption, probability density function, and apriori probability of the targets. It is not easy to apply. The Backpropagation Model 3 was selected from the field of neural network technology, It is a three layered network; input, hidden and output layers. For target identification problem, one has to train the network with the known target (X), then apply the unknown target(Y) to the trained network as an input layer, if the output layer has a higher energy value , thentecWe use two published 4 numerical dat

Report developed under STTR Contract for topic ARMY 02-T004 Under this effort, Frontier Technology, Inc. (FTI) and University of Florida (UF) are developing designs for automatically-generated statistical pattern recognition systems (GASPs) that can classify uncooperative targets among time-varying natural and manmade backgrounds. We also propose to analyze the performance of the envisioned GASPs to: (a) covertly acquire feature data (e.g., statistical, spectral, and spatial cues) from target/background imagery, (b)apply multiple classifiers to target(background information to select probable target location and identity, (c) apply inferencing rules to disambiguate infeasible or contradictory classifier outputs. Pattern selection, key to successful system operation in mission- and threat-specific scenarios, will utilize Dempster-Schaefer theory and UF's powerful data fusion paradigm, Morphological Neural Nets (MNN). Phase-I will evaluate, extend and exploit FTI and UF's successful, DoD-sponsored R&D for dynamic pattern recognition and ATR to develop and test an efficient system design for target classifier output fusion and disambiguation. System design will include analysis of complexity and cost of potential hardware implementations. In a Phase-II effort, we will use Phase-I results to drive candidate pattern downselection in FTI's DoD-supported TNE paradigm. MNNs and TNE have been proven highly successful in a wide variety of recognition problems, thus we propose to analyze GASP system performance in realistic ATR scenarios.

During the last two decades, statistical pattern recognition was well developed in theory and applications with the peak activity in the late sixties. The paper outlines important but unsolved problem areas in statistical pattern recognition and then takes a new and close look at some problems which are related to the finite sample size constraint. In an effort to bridge the gap between theory and practice, constructive solutions are provided for the problems: finite sample distance and information measures, finite sample nearest neighbor decision rule, contextual analysis, decision rules based on discrete and continuous measurements, and finite sample stochastic syntax analysis. It is concluded that there are still many challenging problems to be solved in statistical pattern recognition and every effort should be made such that the theory works well in practice.

During the last two decades, statistical pattern recognition was well developed in theory and applications with the peak activity in the late sixties. The paper outlines important but unsolved problem areas in statistical pattern recognition and then takes a new and close look at some problems which are related to the finite sample size constraint. In an effort to bridge the gap between theory and practice, constructive solutions are provided for the problems: finite sample distance and information measures, finite sample nearest neighbor decision rule, contextual analysis, decision rules based on discrete and continuous measurements, and finite sample stochastic syntax analysis. It is concluded that there are still many challenging problems to be solved in statistical pattern recognition and every effort should be made such that the theory works well in practice.

This report discusses statistical analyses of the spatial pattern of metallic anomalies, buried and on the ground surface, detected during airborne surveys above two former Air Force bombing ranges: the former Pueblo Precision Bombing Range Number 2 in Otero County, Colorado, and the Victorville Precision Bombing Range in San Bernardino County, California. The main purpose of the analyses is to determine whether statistical properties of anomaly spatial patterns can be used to delineate areas with and without unexploded ordnance (UXO). A second goal is to estimate the expected number of UXO at given locations within former military sites based on the results of airborne surveys. Whereas previous statistical characterization of metallic anomaly data at UXO sites has used geostatistical approaches that convert discrete anomaly locations to concentrations of UXO per unit area (McKenna et al. 2002; McKenna and Saito 2003; Saito et al. 2005a,b), this paper focuses on analytical methods that preserve the discrete nature of the pattern information that the data contain. This research is part of the UXO Wide Area Assessment program run by the Department of Defense (DOD) Environmental Security Technology Certification Program (ESTCP). The goal of this program is to provide the DOD with the tools needed to fully characterize the amount of land in the United States that is contaminated with UXO. Due to the long time period during which the military has conducted live-fire training in some areas of the United States, combined with poor record-keeping over much of the DOD s history, the amount of land contaminated with UXO and specific geographic boundaries of the contamination are highly uncertain (Defense Science Board 2003).

The rule space model permits measurement of cognitive skill acquisition, diagnosis of cognitive errors, and detection of the strengths and weaknesses of knowledge possessed by individuals. Two ways to classify an individual into his or her most plausible latent state of knowledge include: (1) hypothesis testing--Bayes' decision rules for minimum errors; and (2) bug distribution--how bugs, incorrect rules used to solve problems, are clustered and related. A 40-item test containing subtraction of fraction items was given to 535 junior high school students. A computer program was used on the PLATO system to diagnose erroneous rules of operation. Two common erroneous problem-solving rules were used to illustrate the rule space model. The results were then compared with the results obtained from a conventional artificial intelligence approach. (GDC)

patternrecogniti0000robe OpenLibrary:- OL3353071W

State-of-the-art research on spectral estimation, feature extraction, and pattern recognition algorithms are presented for radar signal processing and automatic target recognition. Advanced space-time spectral estimation algorithms are presented for multiple moving target feature extraction as well as clutter and jamming suppression for airborne high range resolution (HRR) phased-array radar. A nonparametric adaptive filtering-based approach, referred to as the Gapped-data Amplitude and Phase EStimation (GAPES) algorithm, is proposed for the spectral analysis of gapped data sequences as well as synthetic aperture radar (SAR) imaging with angle diversity data fusion. A QUasi-parametric ALgorithm for target feature Extraction (QUALE) algorithm is also investigated for angle diversity data fusion. Support Vector Machines (SVMs) as compared with other advanced classifiers in the MSTAR Public Domain Release and HRR data are found to outperform neural networks and matched filters. A new concept to create negative examples from the known target class is presented and shown to tremendously improve the rejection of confusers. Finally, Information Theoretic Learning (ITL) is proposed as a new algorithm to demix HRR signatures of closely parked targets.

This paper introduces a probabilistic model that is capable of diagnosing and classifying cognitive errors in a general problem-solving domain. The model is different from the usual deterministic strategies common in the area of artificial intelligence because the item response theory is utilized for handling the variability of response errors. As for illustrating the model, the dataset obtained form a 38-item fraction addition test is used, and the students' responses are classified into 34 groups of misconceptions. These groups are predetermined by the result of an error analysis previously done, and validated with the error diagnostic program written by a typical formal logic approach. Keywords: cognitive errors, item response theory, bugs, fractions, pattern classification, caution index.

Classic statistical techniques (like the multi-dimensional likelihood and the Fisher discriminant method) together with Multi-layer Perceptron and Learning Vector Quantization Neural Networks have been systematically used in order to find the best sensitivity when searching for $\nu_\mu \to \nu_{\tau}$ oscillations. We discovered that for a general direct $\nu_\tau$ appearance search based on kinematic criteria: a) An optimal discrimination power is obtained using only three variables ($E_{visible}$, $P_{T}^{miss}$ and $\rho_{l}$) and their correlations. Increasing the number of variables (or combinations of variables) only increases the complexity of the problem, but does not result in a sensible change of the expected sensitivity. b) The multi-layer perceptron approach offers the best performance. As an example to assert numerically those points, we have considered the problem of $\nu_\tau$ appearance at the CNGS beam using a Liquid Argon TPC detector.

Statistical methods for pattern recognition and classification applications

Includes bibliographical references and index

In this paper, we compare different existing approaches employed in data mining of big proof libraries in automated and interactive theorem proving.