J-PMCRI: a methodology for inducing pre-pruned modular classification rules

Inducing rules from very large datasets is one of the most challenging areas in data mining. Several approaches exist to scaling up classification rule induction to large datasets, namely data reduction and the parallelisation of classification rule induction algorithms. In the area of parallelisation of classification rule induction algorithms most of the work has been concentrated on the Top Down Induction of Decision Trees (TDIDT), also known as the ‘divide and conquer’ approach. However powerful alternative algorithms exist that induce modular rules. Most of these alternative algorithms follow the ‘separate and conquer’ approach of inducing rules, but very little work has been done to make the ‘separate and conquer’ approach scale better on large training data. This paper examines the potential of the recently developed blackboard based J-PMCRI methodology for parallelising modular classification rule induction algorithms that follow the ‘separate and conquer’ approach. A concrete implementation of the methodology is evaluated empirically on very large datasets.

Parallel random prism: a computationally efficient ensemble learner for classification

Generally classifiers tend to overfit if there is noise in the training data or there are missing values. Ensemble learning methods are often used to improve a classifier's classification accuracy. Most ensemble learning approaches aim to improve the classification accuracy of decision trees. However, alternative classifiers to decision trees exist. The recently developed Random Prism ensemble learner for classification aims to improve an alternative classification rule induction approach, the Prism family of algorithms, which addresses some of the limitations of decision trees. However, Random Prism suffers like any ensemble learner from a high computational overhead due to replication of the data and the induction of multiple base classifiers. Hence even modest sized datasets may impose a computational challenge to ensemble learners such as Random Prism. Parallelism is often used to scale up algorithms to deal with large datasets. This paper investigates parallelisation for Random Prism, implements a prototype and evaluates it empirically using a Hadoop computing cluster.

J-measure based hybrid pruning for complexity reduction in classification rules

Prism is a modular classification rule generation method based on the ‘separate and conquer’ approach that is alternative to the rule induction approach using decision trees also known as ‘divide and conquer’. Prism often achieves a similar level of classification accuracy compared with decision trees, but tends to produce a more compact noise tolerant set of classification rules. As with other classification rule generation methods, a principle problem arising with Prism is that of overfitting due to over-specialised rules. In addition, over-specialised rules increase the associated computational complexity. These problems can be solved by pruning methods. For the Prism method, two pruning algorithms have been introduced recently for reducing overfitting of classification rules - J-pruning and Jmax-pruning. Both algorithms are based on the J-measure, an information theoretic means for quantifying the theoretical information content of a rule. Jmax-pruning attempts to exploit the J-measure to its full potential because J-pruning does not actually achieve this and may even lead to underfitting. A series of experiments have proved that Jmax-pruning may outperform J-pruning in reducing overfitting. However, Jmax-pruning is computationally relatively expensive and may also lead to underfitting. This paper reviews the Prism method and the two existing pruning algorithms above. It also proposes a novel pruning algorithm called Jmid-pruning. The latter is based on the J-measure and it reduces overfitting to a similar level as the other two algorithms but is better in avoiding underfitting and unnecessary computational effort. The authors conduct an experimental study on the performance of the Jmid-pruning algorithm in terms of classification accuracy and computational efficiency. The algorithm is also evaluated comparatively with the J-pruning and Jmax-pruning algorithms.

A hybrid approach to learn with imbalanced classes using evolutionary algorithms

There is an increasing interest in the application of Evolutionary Algorithms (EAs) to induce classification rules. This hybrid approach can benefit areas where classical methods for rule induction have not been very successful. One example is the induction of classification rules in imbalanced domains. Imbalanced data occur when one or more classes heavily outnumber other classes. Frequently, classical machine learning (ML) classifiers are not able to learn in the presence of imbalanced data sets, inducing classification models that always predict the most numerous classes. In this work, we propose a novel hybrid approach to deal with this problem. We create several balanced data sets with all minority class cases and a random sample of majority class cases. These balanced data sets are fed to classical ML systems that produce rule sets. The rule sets are combined creating a pool of rules and an EA is used to build a classifier from this pool of rules. This hybrid approach has some advantages over undersampling, since it reduces the amount of discarded information, and some advantages over oversampling, since it avoids overfitting. The proposed approach was experimentally analysed and the experimental results show an improvement in the classification performance measured as the area under the receiver operating characteristics (ROC) curve.

Classification of petroleum well drilling operations with a hybrid particle swarm/ant colony algorithm

This paper describes an investigation of the hybrid PSO/ACO algorithm to classify automatically the well drilling operation stages. The method feasibility is demonstrated by its application to real mud-logging dataset. The results are compared with bio-inspired methods, and rule induction and decision tree algorithms for data mining. © 2009 Springer Berlin Heidelberg.

Semantic Service Discovery Techniques for the composable web

This PhD thesis contributes to the problem of resource and service discovery in the context of the composable web. In the current web, mashup technologies allow developers reusing services and contents to build new web applications. However, developers face a problem of information flood when searching for appropriate services or resources for their combination. To contribute to overcoming this problem, a framework is defined for the discovery of services and resources. In this framework, three levels are defined for performing discovery at content, discovery and agente levels. The content level involves the information available in web resources. The web follows the Representational Stateless Transfer (REST) architectural style, in which resources are returned as representations from servers to clients. These representations usually employ the HyperText Markup Language (HTML), which, along with Content Style Sheets (CSS), describes the markup employed to render representations in a web browser. Although the use of SemanticWeb standards such as Resource Description Framework (RDF) make this architecture suitable for automatic processes to use the information present in web resources, these standards are too often not employed, so automation must rely on processing HTML. This process, often referred as Screen Scraping in the literature, is the content discovery according to the proposed framework. At this level, discovery rules indicate how the different pieces of data in resources’ representations are mapped onto semantic entities. By processing discovery rules on web resources, semantically described contents can be obtained out of them. The service level involves the operations that can be performed on the web. The current web allows users to perform different tasks such as search, blogging, e-commerce, or social networking. To describe the possible services in RESTful architectures, a high-level feature-oriented service methodology is proposed at this level. This lightweight description framework allows defining service discovery rules to identify operations in interactions with REST resources. The discovery is thus performed by applying discovery rules to contents discovered in REST interactions, in a novel process called service probing. Also, service discovery can be performed by modelling services as contents, i.e., by retrieving Application Programming Interface (API) documentation and API listings in service registries such as ProgrammableWeb. For this, a unified model for composable components in Mashup-Driven Development (MDD) has been defined after the analysis of service repositories from the web. The agent level involves the orchestration of the discovery of services and contents. At this level, agent rules allow to specify behaviours for crawling and executing services, which results in the fulfilment of a high-level goal. Agent rules are plans that allow introspecting the discovered data and services from the web and the knowledge present in service and content discovery rules to anticipate the contents and services to be found on specific resources from the web. By the definition of plans, an agent can be configured to target specific resources. The discovery framework has been evaluated on different scenarios, each one covering different levels of the framework. Contenidos a la Carta project deals with the mashing-up of news from electronic newspapers, and the framework was used for the discovery and extraction of pieces of news from the web. Similarly, in Resulta and VulneraNET projects the discovery of ideas and security knowledge in the web is covered, respectively. The service level is covered in the OMELETTE project, where mashup components such as services and widgets are discovered from component repositories from the web. The agent level is applied to the crawling of services and news in these scenarios, highlighting how the semantic description of rules and extracted data can provide complex behaviours and orchestrations of tasks in the web. The main contributions of the thesis are the unified framework for discovery, which allows configuring agents to perform automated tasks. Also, a scraping ontology has been defined for the construction of mappings for scraping web resources. A novel first-order logic rule induction algorithm is defined for the automated construction and maintenance of these mappings out of the visual information in web resources. Additionally, a common unified model for the discovery of services is defined, which allows sharing service descriptions. Future work comprises the further extension of service probing, resource ranking, the extension of the Scraping Ontology, extensions of the agent model, and contructing a base of discovery rules. Resumen La presente tesis doctoral contribuye al problema de descubrimiento de servicios y recursos en el contexto de la web combinable. En la web actual, las tecnologías de combinación de aplicaciones permiten a los desarrolladores reutilizar servicios y contenidos para construir nuevas aplicaciones web. Pese a todo, los desarrolladores afrontan un problema de saturación de información a la hora de buscar servicios o recursos apropiados para su combinación. Para contribuir a la solución de este problema, se propone un marco de trabajo para el descubrimiento de servicios y recursos. En este marco, se definen tres capas sobre las que se realiza descubrimiento a nivel de contenido, servicio y agente. El nivel de contenido involucra a la información disponible en recursos web. La web sigue el estilo arquitectónico Representational Stateless Transfer (REST), en el que los recursos son devueltos como representaciones por parte de los servidores a los clientes. Estas representaciones normalmente emplean el lenguaje de marcado HyperText Markup Language (HTML), que, unido al estándar Content Style Sheets (CSS), describe el marcado empleado para mostrar representaciones en un navegador web. Aunque el uso de estándares de la web semántica como Resource Description Framework (RDF) hace apta esta arquitectura para su uso por procesos automatizados, estos estándares no son empleados en muchas ocasiones, por lo que cualquier automatización debe basarse en el procesado del marcado HTML. Este proceso, normalmente conocido como Screen Scraping en la literatura, es el descubrimiento de contenidos en el marco de trabajo propuesto. En este nivel, un conjunto de reglas de descubrimiento indican cómo los diferentes datos en las representaciones de recursos se corresponden con entidades semánticas. Al procesar estas reglas sobre recursos web, pueden obtenerse contenidos descritos semánticamente. El nivel de servicio involucra las operaciones que pueden ser llevadas a cabo en la web. Actualmente, los usuarios de la web pueden realizar diversas tareas como búsqueda, blogging, comercio electrónico o redes sociales. Para describir los posibles servicios en arquitecturas REST, se propone en este nivel una metodología de alto nivel para descubrimiento de servicios orientada a funcionalidades. Este marco de descubrimiento ligero permite definir reglas de descubrimiento de servicios para identificar operaciones en interacciones con recursos REST. Este descubrimiento es por tanto llevado a cabo al aplicar las reglas de descubrimiento sobre contenidos descubiertos en interacciones REST, en un nuevo procedimiento llamado sondeo de servicios. Además, el descubrimiento de servicios puede ser llevado a cabo mediante el modelado de servicios como contenidos. Es decir, mediante la recuperación de documentación de Application Programming Interfaces (APIs) y listas de APIs en registros de servicios como ProgrammableWeb. Para ello, se ha definido un modelo unificado de componentes combinables para Mashup-Driven Development (MDD) tras el análisis de repositorios de servicios de la web. El nivel de agente involucra la orquestación del descubrimiento de servicios y contenidos. En este nivel, las reglas de nivel de agente permiten especificar comportamientos para el rastreo y ejecución de servicios, lo que permite la consecución de metas de mayor nivel. Las reglas de los agentes son planes que permiten la introspección sobre los datos y servicios descubiertos, así como sobre el conocimiento presente en las reglas de descubrimiento de servicios y contenidos para anticipar contenidos y servicios por encontrar en recursos específicos de la web. Mediante la definición de planes, un agente puede ser configurado para descubrir recursos específicos. El marco de descubrimiento ha sido evaluado sobre diferentes escenarios, cada uno cubriendo distintos niveles del marco. El proyecto Contenidos a la Carta trata de la combinación de noticias de periódicos digitales, y en él el framework se ha empleado para el descubrimiento y extracción de noticias de la web. De manera análoga, en los proyectos Resulta y VulneraNET se ha llevado a cabo un descubrimiento de ideas y de conocimientos de seguridad, respectivamente. El nivel de servicio se cubre en el proyecto OMELETTE, en el que componentes combinables como servicios y widgets se descubren en repositorios de componentes de la web. El nivel de agente se aplica al rastreo de servicios y noticias en estos escenarios, mostrando cómo la descripción semántica de reglas y datos extraídos permiten proporcionar comportamientos complejos y orquestaciones de tareas en la web. Las principales contribuciones de la tesis son el marco de trabajo unificado para descubrimiento, que permite configurar agentes para realizar tareas automatizadas. Además, una ontología de extracción ha sido definida para la construcción de correspondencias y extraer información de recursos web. Asimismo, un algoritmo para la inducción de reglas de lógica de primer orden se ha definido para la construcción y el mantenimiento de estas correspondencias a partir de la información visual de recursos web. Adicionalmente, se ha definido un modelo común y unificado para el descubrimiento de servicios que permite la compartición de descripciones de servicios. Como trabajos futuros se considera la extensión del sondeo de servicios, clasificación de recursos, extensión de la ontología de extracción y la construcción de una base de reglas de descubrimiento.

Crowbar System in Doubly Fed Induction Wind Generators

In the last 15 years, the use of doubly fed induction machines in modern variable-speed wind turbines has increased rapidly. This development has been driven by the cost reduction as well as the low-loss generation of Insulated Gate Bipolar Transistors (IGBT). According to new grid code requirements, wind turbines must remain connected to the grid during grid disturbances. Moreover, they must also contribute to voltage support during and after grid faults. The crowbar system is essential to avoid the disconnection of the doubly fed induction wind generators from the network during faults. The insertion of the crowbar in the rotor circuits for a short period of time enables a more efficient terminal voltage control. As a general rule, the activation and the deactivation of the crowbar system is based only on the DC-link voltage level of the back-to-back converters. In this context, the authors discuss the critical rotor speed to analyze the instability of doubly fed induction generators during grid faults.

Induction of modular classification rules: using Jmax-pruning

The Prism family of algorithms induces modular classification rules which, in contrast to decision tree induction algorithms, do not necessarily fit together into a decision tree structure. Classifiers induced by Prism algorithms achieve a comparable accuracy compared with decision trees and in some cases even outperform decision trees. Both kinds of algorithms tend to overfit on large and noisy datasets and this has led to the development of pruning methods. Pruning methods use various metrics to truncate decision trees or to eliminate whole rules or single rule terms from a Prism rule set. For decision trees many pre-pruning and postpruning methods exist, however for Prism algorithms only one pre-pruning method has been developed, J-pruning. Recent work with Prism algorithms examined J-pruning in the context of very large datasets and found that the current method does not use its full potential. This paper revisits the J-pruning method for the Prism family of algorithms and develops a new pruning method Jmax-pruning, discusses it in theoretical terms and evaluates it empirically.

Automatic induction of lexical features

This thesis concerns artificially intelligent natural language processing systems that are capable of learning the properties of lexical items (properties like verbal valency or inflectional class membership) autonomously while they are fulfilling their tasks for which they have been deployed in the first place. Many of these tasks require a deep analysis of language input, which can be characterized as a mapping of utterances in a given input C to a set S of linguistically motivated structures with the help of linguistic information encoded in a grammar G and a lexicon L: G + L + C → S (1) The idea that underlies intelligent lexical acquisition systems is to modify this schematic formula in such a way that the system is able to exploit the information encoded in S to create a new, improved version of the lexicon: G + L + S → L' (2) Moreover, the thesis claims that a system can only be considered intelligent if it does not just make maximum usage of the learning opportunities in C, but if it is also able to revise falsely acquired lexical knowledge. So, one of the central elements in this work is the formulation of a couple of criteria for intelligent lexical acquisition systems subsumed under one paradigm: the Learn-Alpha design rule. The thesis describes the design and quality of a prototype for such a system, whose acquisition components have been developed from scratch and built on top of one of the state-of-the-art Head-driven Phrase Structure Grammar (HPSG) processing systems. The quality of this prototype is investigated in a series of experiments, in which the system is fed with extracts of a large English corpus. While the idea of using machine-readable language input to automatically acquire lexical knowledge is not new, we are not aware of a system that fulfills Learn-Alpha and is able to deal with large corpora. To instance four major challenges of constructing such a system, it should be mentioned that a) the high number of possible structural descriptions caused by highly underspeci ed lexical entries demands for a parser with a very effective ambiguity management system, b) the automatic construction of concise lexical entries out of a bulk of observed lexical facts requires a special technique of data alignment, c) the reliability of these entries depends on the system's decision on whether it has seen 'enough' input and d) general properties of language might render some lexical features indeterminable if the system tries to acquire them with a too high precision. The cornerstone of this dissertation is the motivation and development of a general theory of automatic lexical acquisition that is applicable to every language and independent of any particular theory of grammar or lexicon. This work is divided into five chapters. The introductory chapter first contrasts three different and mutually incompatible approaches to (artificial) lexical acquisition: cue-based queries, head-lexicalized probabilistic context free grammars and learning by unification. Then the postulation of the Learn-Alpha design rule is presented. The second chapter outlines the theory that underlies Learn-Alpha and exposes all the related notions and concepts required for a proper understanding of artificial lexical acquisition. Chapter 3 develops the prototyped acquisition method, called ANALYZE-LEARN-REDUCE, a framework which implements Learn-Alpha. The fourth chapter presents the design and results of a bootstrapping experiment conducted on this prototype: lexeme detection, learning of verbal valency, categorization into nominal count/mass classes, selection of prepositions and sentential complements, among others. The thesis concludes with a review of the conclusions and motivation for further improvements as well as proposals for future research on the automatic induction of lexical features.

Evidence of a transition from perceptual to category induction in 3- to 9-year-old children

We examined whether inductive reasoning development is better characterized by accounts assuming an early category bias versus an early perceptual bias. We trained 264 children aged 3 to 9 years to categorize novel insects using a rule that directly pitted category membership against appearance. This was followed by an induction task with perceptual distractors at different levels of featural similarity. An additional 52 children were given the same training followed by an induction task with alternative stimuli. Categorization performance was consistently high, however we found a gradual transition from a perceptual bias in our youngest children to a category bias around age 6-7. In addition, children of all ages were equally distracted by higher levels of featural similarity. The transition is unlikely to be due to an increased ability to inhibit perceptual distractors. Instead, we argue that the transition is driven by a fundamental change in children’s understanding of category membership.

Ikkε Is Key To Induction Of Insulin Resistance In The Hypothalamus, And Its Inhibition Reverses Obesity.

IKK epsilon (IKKε) is induced by the activation of nuclear factor-κB (NF-κB). Whole-body IKKε knockout mice on a high-fat diet (HFD) were protected from insulin resistance and showed altered energy balance. We demonstrate that IKKε is expressed in neurons and is upregulated in the hypothalamus of obese mice, contributing to insulin and leptin resistance. Blocking IKKε in the hypothalamus of obese mice with CAYMAN10576 or small interfering RNA decreased NF-κB activation in this tissue, relieving the inflammatory environment. Inhibition of IKKε activity, but not TBK1, reduced IRS-1(Ser307) phosphorylation and insulin and leptin resistance by an improvement of the IR/IRS-1/Akt and JAK2/STAT3 pathways in the hypothalamus. These improvements were independent of body weight and food intake. Increased insulin and leptin action/signaling in the hypothalamus may contribute to a decrease in adiposity and hypophagia and an enhancement of energy expenditure accompanied by lower NPY and increased POMC mRNA levels. Improvement of hypothalamic insulin action decreases fasting glycemia, glycemia after pyruvate injection, and PEPCK protein expression in the liver of HFD-fed and db/db mice, suggesting a reduction in hepatic glucose production. We suggest that IKKε may be a key inflammatory mediator in the hypothalamus of obese mice, and its hypothalamic inhibition improves energy and glucose metabolism.

[comparative Study Between Fast And Slow Induction Of Propofol Given By Target-controlled Infusion: Expected Propofol Concentration At The Effect Site. Randomized Controlled Trial].

studies have shown that rate of propofol infusion may influence the predicted propofol concentration at the effect site (Es). The aim of this study was to evaluate the Es predicted by the Marsh pharmacokinetic model (ke0 0.26min(-1)) in loss of consciousness during fast or slow induction. the study included 28 patients randomly divided into two equal groups. In slow induction group (S), target-controlled infusion (TCI) of propofol with plasma, Marsh pharmacokinetic model (ke0 0.26min(-1)) with target concentration (Tc) at 2.0-μg.mL(-1) were administered. When the predicted propofol concentration at the effect site (Es) reached half of Es value, Es was increased to previous Es + 1μg.mL(-1), successively, until loss of consciousness. In rapid induction group (R), patients were induced with TCI of propofol with plasma (6.0μg.ml(-1)) at Es, and waited until loss of consciousness. in rapid induction group, Tc for loss of consciousness was significantly lower compared to slow induction group (1.67±0.76 and 2.50±0.56μg.mL(-1), respectively, p=0.004). the predicted propofol concentration at the effect site for loss of consciousness is different for rapid induction and slow induction, even with the same pharmacokinetic model of propofol and the same balance constant between plasma and effect site.

Quando o estado joga a favor do privado : as politicas de esporte apos a constituição de 1988

Universidade Estadual de Campinas . Faculdade de Educação Física

Histopathological evaluation of different methods of experimental induction of periapical periodontitis

This study evaluated histopathologically different methods of experimental induction of periapical periodontitis. The radiographic and microbiological evaluations have been performed in a previous investigation. Fifty-seven root canals from dogs' teeth were assigned to 4 groups. In GI (n=14) and GII (n=14), the root canals were exposed to oral environment for 180 days; in GIII (n=14) and GIV (n=15) the root canals were exposed for 7 days and then the access cavities were restored and remained sealed for 53 days. The root apices of GI and GIII were perforated, whilst those of GII and GIV remained intact. After induction of periapical periodontitis, the dogs were euthanized. Serial sections were obtained and stained with hematoxylin and eosin. Data of the histopathological evaluation were submitted to Kruskal-Wallis and Dunn's tests at 5% significance level. The inflammatory periapical reaction and resorption of mineralized tissues were less intense in GII than in the other groups (p<0.05). There was no histopathological difference among the experimentally induced periapical lesions in the teeth with coronal sealing. On the other hand, when coronal sealing was not performed, greater intensity of induced periapical periodontitis was observed in the teeth with apical perforation.

Microbial distribution in the root canal system after periapical lesion induction using different methods

The aim of this study was to evaluate the microbial distribution in the root canal system after periapical lesion induction in dogs' teeth using different methods. Fifty-two root canals were assigned to 4 groups (n=13). Groups I and II: root canals were exposed to the oral cavity for 180 days; groups III and IV: root canals were exposed for 7 days and then the coronal openings were sealed for 53 days. The root apices of groups I and III were perforated, while those of groups II and IV remained intact. After the experimental periods, the animals were euthanized and the anatomic pieces containing the roots were processed and stained with the Brown & Brenn method to assess the presence and distribution of microorganisms. The incidence of microorganisms at different sites of the roots and periapical lesions was analyzed statistically by the chi-square test at 5% significance level. All groups presented microorganisms in the entire root canal system. A larger number of microorganisms was observed on the root canal walls, apical delta and dentinal tubules (p<0.05), followed by cementum and cemental resorption areas. In spite of the different periods of exposure to the oral environment, the methods used for induction of periapical periodontitis yielded similar distribution of microorganisms in the root canal system.