Search

Problems with large monolithical ontologies in terms of reusability, scalability and maintenance have led to an increasing interest in modularization techniques for ontologies. Currently, existing work suffers from the fact that the notion of modularization is not as well understood in the context of ontologies as it is in software engineering. In this paper, we experiment on applying state-of-the-art tools for ontology modularization in the context of a concrete application: the automatic selection of knowledge components to be used for Web page annotation and semantic browsing. We conclude that, in a broader context, an evaluation framework is required to guide the choice of a modularization tool, in accordance with the requirements of the considered application. [ABSTRACT FROM AUTHOR]

In this work, we proposed the use of Support Vector Machines (SVM) to predict the performance of machine learning algorithms based on features of the learning problems. This work is related to the Meta-Regression approach, which has been successfully applied to predict learning performance, supporting algorithm selection. Experiments were performed in a case study in which SVMs with different kernel functions were used to predict the performance of Multi-Layer Perceptron (MLP) networks. The SVMs obtained better results in the evaluated task, when compared to different algorithms that have been applied as meta-regressors in previous work. [ABSTRACT FROM AUTHOR]

This paper presents a context driven segmentation and recognition method for handwritten Chinese characters. We follow a split-merge technique in character segmentation. In this process, a Chinese text line is first pre-segmented into a sequence of radicals, which are then merged according to a cost function combining both recognition confidence and contextual cost. Two strategies are also proposed for implementation: bi-gram based merging and lexicon driven merging. In the former one, we generate a set of merging paths which are then evaluated by Viterbi algorithm. The radicals' best merging method is given by the path with the highest score. In the latter strategy, a lexicon is preset and compared with the radicals to determine both radicals' merging and candidate character selection. Experiments show that contextual information plays a crucial role in Chinese character segmentation and could obviously improve the segmentation and recognition results. [ABSTRACT FROM AUTHOR]

An on-line signature verification system exploiting both local and global information through decision-level fusion is presented. Global information is extracted with a feature-based representation and recognized by using Parzen Windows Classifiers. Local information is extracted as time functions of various dynamic properties and recognized by using Hidden Markov Models. Experimental results are given on the large MCYT signature database (330 signers, 16500 signatures) for random and skilled forgeries. Feature selection experiments based on feature ranking are carried out. It is shown experimentally that the machine expert based on local information outperforms the system based on global analysis when enough training data is available. Conversely, it is found that global analysis is more appropriate in the case of small training set size. The two proposed systems are also shown to give complementary recognition information which is successfully exploited using decision-level score fusion. [ABSTRACT FROM AUTHOR]

This paper presents a novel method for feature extraction based on the generalized entropy of the histogram formed by Euclidean distances, which is named distributive entropy of Euclidean distance (DEED in sort). DEED is a nonlinear measure for learning feature space, which provides the congregate and information measure of learning samples space. The ratio of between-class DEED to within-class DEED (Jrd) is used as a new nonlinear separability criterion for optimizing feature selection. Experiments on vehicle classification show that the proposed method has better performance on all the datasets than the fisher linear discriminant analysis. [ABSTRACT FROM AUTHOR]

Host specificity is arguably one of the most important properties of a parasite, because it can determine, among other things, whether a parasite can survive the extinction of a host species, whether a parasite has the potential to invade new habitats such as islands, or whether a parasite can become established and spread following its introduction to a new geographical area. Macroparasites of micromammals have received relatively little attention in this regard. The available evidence suggests that some ecological features of small mammals may interact with parasite transmission mode to determine what levels of host specificity are observed. Still, large-scale patterns of host specificity have only been investigated in fleas, and studies on other parasite taxa are definitely needed. In addition, since many rodent species are now universal laboratory models in many branches of biology, it should prove possible to investigate host specificity in an experimental context. For instance, the mechanisms responsible for the failure or success of a particular parasite species in different host species could be examined using controlled laboratory infections. In addition, selection experiments like that of Gemmill et al. (2000) can be envisaged with host species like mice with short generation times, to track the evolution of host specificity under different selection regimes. The evolution and ecology of host specificity will remain an important research area for years to come. This is particularly true in the light of the global environmental changes occurring at present, and the possibility that, by altering transmission conditions, they will lead to the expansion of the host range of many parasite species. [ABSTRACT FROM AUTHOR]

Gene Selection is one class of most used data analysis algorithms on microarray dataset. The goal of gene selection algorithms is to filter out a small set of informative genes that best explains experimental variations. Traditional gene selection algorithms are mostly single-gene based. Some discriminative scores are calculated and sorted for each gene. Top ranked genes are then selected as informative genes for further study. Such algorithms ignore completely correlations between genes, although such correlations is widely known. Genes interact with each other through various pathways and regulative networks. In this paper, we propose to use, instead of ignoring, such correlations for gene selection. Experiments performed on three public available datasets show promising results. [ABSTRACT FROM AUTHOR]

This paper describes a document classifier for web documents in the fields of Information Technology and uses SVM to learn a model, which is constructed from the training sets and its representative terms. To reduce information overload, it needs to exploit automatic text classification for handling enormous documents. Support Vector Machine (SVM) is a model that is calculated as a weighted sum of kernel function outputs. The basic idea is to exploit the representative terms meaning distribution in coherent thematic texts of each category by simple statistics methods. Vector-space model is applied to represent documents in the categories by using feature selection scheme based on TFiDF. We apply a category factor which represents effects in category of any term to the feature selection. Experiments show the results of categorization and the correlation of vector length. [ABSTRACT FROM AUTHOR]

Gene Selection is one class of most used data analysis algorithms on microarray datasets. The goal of gene selection algorithms is to filter out a small set of informative genes that best explains experimental variations. Traditional gene selection algorithms are mostly single-gene based. Some discriminative scores are calculated and sorted for each gene. Top ranked genes are then selected as informative genes for further study. Such algorithms ignore completely correlations between genes, although such correlations is widely known. Genes interact with each other through various pathways and regulative networks. In this paper, we propose to use, instead of ignoring, such correlations for gene selection. Experiments performed on three public available datasets show promising results. [ABSTRACT FROM AUTHOR]

Structural variation in organisms has been a topic of long-standing interest in biology and attempts to account for and explain this variation were major contributors to the formulation of Darwin's ideas about evolution. Structural variation is the currency of taxonomy, systematics, functional ecology, and much of evolutionary biology. Whereas it is assumed that changes in structure are related to changes in function, and vice versa, teasing out the intricate relationships of this couplet is a perplexing task. The relationship of structure to function at the intraspecific level has been investigated chiefly in two distinct, but related ways. In situ studies have posed questions about variation as seen in the field at various scales, from the very local to those encompassing large geographic distances. In this overall approach, structural variation has been correlated, in a variety of ways, with environmental variables, both biotic and abiotic. Relying on various theoretical underpinnings of why and how such variation occurs, in situ studies have investigated variation in the context of trophic polymorphism within and between populations, clinal variation across broad geographic ranges, and microgeographic variation across smaller ranges. Ex situ studies have largely examined structural variation in an experimental, laboratory-type context and have focused on performance comparisons between exemplars exhibiting observed structural variations. Such approaches have considered variation in trophic and, more extensively, locomotor performance and attempt to explain them by way of variation in measured performance. Fluctuating asymmetry has recently become of increasing interest in terms of its influence on performance, as have selection experiments that attempt to explore the limits of variation in relation to performance. There is an increasing tendency to combine in situ and ex situ components in integrated investigations, but the scale and intensity of such undertakings present new challenges, and also reveal new intricacies and revelations that challenge the extrapolations that tend to be made from single approaches in isolation. Our ability to explain the relationship between structure and function as they relate to variation may be dependent upon the level of detail we expect from the explanatory model. [ABSTRACT FROM AUTHOR]