Particle competition and cooperation to prevent error propagation from mislabeled data in semi-supervised learning

Semi-supervised learning is applied to classification problems where only a small portion of the data items is labeled. In these cases, the reliability of the labels is a crucial factor, because mislabeled items may propagate wrong labels to a large portion or even the entire data set. This paper aims to address this problem by presenting a graph-based (network-based) semi-supervised learning method, specifically designed to handle data sets with mislabeled samples. The method uses teams of walking particles, with competitive and cooperative behavior, for label propagation in the network constructed from the input data set. The proposed model is nature-inspired and it incorporates some features to make it robust to a considerable amount of mislabeled data items. Computer simulations show the performance of the method in the presence of different percentage of mislabeled data, in networks of different sizes and average node degree. Importantly, these simulations reveals the existence of the critical points of the mislabeled subset size, below which the network is free of wrong label contamination, but above which the mislabeled samples start to propagate their labels to the rest of the network. Moreover, numerical comparisons have been made among the proposed method and other representative graph-based semi-supervised learning methods using both artificial and real-world data sets. Interestingly, the proposed method has increasing better performance than the others as the percentage of mislabeled samples is getting larger. © 2012 IEEE.

BBA: A binary bat algorithm for feature selection

Feature selection aims to find the most important information from a given set of features. As this task can be seen as an optimization problem, the combinatorial growth of the possible solutions may be in-viable for a exhaustive search. In this paper we propose a new nature-inspired feature selection technique based on the bats behaviour, which has never been applied to this context so far. The wrapper approach combines the power of exploration of the bats together with the speed of the Optimum-Path Forest classifier to find the set of features that maximizes the accuracy in a validating set. Experiments conducted in five public datasets have demonstrated that the proposed approach can outperform some well-known swarm-based techniques. © 2012 IEEE.

Multilayer perceptron neural networks training through charged system search and its Application for non-technical losses detection

The non-technical loss is not a problem with trivial solution or regional character and its minimization represents the guarantee of investments in product quality and maintenance of power systems, introduced by a competitive environment after the period of privatization in the national scene. In this paper, we show how to improve the training phase of a neural network-based classifier using a recently proposed meta-heuristic technique called Charged System Search, which is based on the interactions between electrically charged particles. The experiments were carried out in the context of non-technical loss in power distribution systems in a dataset obtained from a Brazilian electrical power company, and have demonstrated the robustness of the proposed technique against with several others nature-inspired optimization techniques for training neural networks. Thus, it is possible to improve some applications on Smart Grids. © 2013 IEEE.

BCS: A Binary Cuckoo Search algorithm for feature selection

Feature selection has been actively pursued in the last years, since to find the most discriminative set of features can enhance the recognition rates and also to make feature extraction faster. In this paper, the propose a new feature selection called Binary Cuckoo Search, which is based on the behavior of cuckoo birds. The experiments were carried out in the context of theft detection in power distribution systems in two datasets obtained from a Brazilian electrical power company, and have demonstrated the robustness of the proposed technique against with several others nature-inspired optimization techniques. © 2013 IEEE.

Optimizing feature selection through binary charged system search

Feature selection aims to find the most important information from a given set of features. As this task can be seen as an optimization problem, the combinatorial growth of the possible solutions may be inviable for a exhaustive search. In this paper we propose a new nature-inspired feature selection technique based on the Charged System Search (CSS), which has never been applied to this context so far. The wrapper approach combines the power of exploration of CSS together with the speed of the Optimum-Path Forest classifier to find the set of features that maximizes the accuracy in a validating set. Experiments conducted in four public datasets have demonstrated the validity of the proposed approach can outperform some well-known swarm-based techniques. © 2013 Springer-Verlag.

Combined active and semi-supervised learning using particle walking temporal dynamicscombined active and semi-supervised learning using particle walking temporal dynamics

Both Semi-Supervised Leaning and Active Learning are techniques used when unlabeled data is abundant, but the process of labeling them is expensive and/or time consuming. In this paper, those two machine learning techniques are combined into a single nature-inspired method. It features particles walking on a network built from the data set, using a unique random-greedy rule to select neighbors to visit. The particles, which have both competitive and cooperative behavior, are created on the network as the result of label queries. They may be created as the algorithm executes and only nodes affected by the new particles have to be updated. Therefore, it saves execution time compared to traditional active learning frameworks, in which the learning algorithm has to be executed several times. The data items to be queried are select based on information extracted from the nodes and particles temporal dynamics. Two different rules for queries are explored in this paper, one of them is based on querying by uncertainty approaches and the other is based on data and labeled nodes distribution. Each of them may perform better than the other according to some data sets peculiarities. Experimental results on some real-world data sets are provided, and the proposed method outperforms the semi-supervised learning method, from which it is derived, in all of them.

Genetic Control of Mosquitoes: population suppression strategies

Over the last two decades, morbidity and mortality from malaria and dengue fever among other pathogens are an increasing Public Health problem. The increase in the geographic distribution of vectors is accompanied by the emergence of viruses and diseases in new areas. There are insufficient specific therapeutic drugs available and there are no reliable vaccines for malaria or dengue, although some progress has been achieved, there is still a long way between its development and actual field use. Most mosquito control measures have failed to achieve their goals, mostly because of the mosquito's great reproductive capacity and genomic flexibility. Chemical control is increasingly restricted due to potential human toxicity, mortality in no target organisms, insecticide resistance, and other environmental impacts. Other strategies for mosquito control are desperately needed. The Sterile Insect Technique (SIT) is a species-specific and environmentally benign method for insect population suppression, it is based on mass rearing, radiation mediated sterilization, and release of a large number of male insects. Releasing of Insects carrying a dominant lethal gene (RIDL) offers a solution to many of the drawbacks of traditional SIT that have limited its application in mosquitoes while maintaining its environmentally friendly and species-specific utility. The self-limiting nature of sterile mosquitoes tends to make the issues related to field use of these somewhat less challenging than for self-spreading systems characteristic of population replacement strategies. They also are closer to field use, so might be appropriate to consider first. The prospect of genetic control methods against mosquito vectored human diseases is rapidly becoming a reality, many decisions will need to be made on a national, regional and international level regarding the biosafety, social, cultural and ethical aspects of the use and deployment of these vector control methods.

Supramolekulare Chemie mit Cyclodextrinen

Zusammenfassung Die vorliegende Arbeit gliedert sich in zwei Teilgebiete, die beide unter dem Gesichtspunkt der Supramolekularen Chemie mit Cyclodextrinen bearbeitet wurden. Der erste Teil befasste sich mit der Synthese und Charakterisierung von chemisch modifizierten Cyclodextrinen, die flüssigkristalline Eigenschaften besitzen sollten. Die verwendeten Cyclodextrine wurden mit verschieden substituierten Benzoesäurederivaten umgesetzt und durch polarisationsmikroskopische Untersuchungen auf flüssigkristalline Phasenübergänge hin untersucht. Weitere Untersuchungen erfolgten mittels DSC und Röntgendiffraktometrie.In wie weit die flüssigkristallinen Eigenschaften eine Selbstorganisation zu Überstrukturen induzierten, sollten Untersuchungen als Monoschichten auf der Luft/Wasser-Grenzfläche mittels einer Filmwaage zeigen. Daraus ergaben sich weitergehende Untersuchungen mit der AFM-Technik. Es zeigte sich, dass besonders die quasi symmetrischen, über Esterverknüpfungen modifizierten Cyclodextrine flüssigkristalline Eigenschaften zeigten. Es konnte auch gezeigt werden, dass sie sich in Mono- und Multischichten zu Lipid-Strukturen organisieren. Der zweite Teil dieser Arbeit befasste sich mit der Wechselwirkung von Cyclodextrinen mit LCST-Copolymeren, die supramolekulare Komplexe bilden. Als Schlüsselgruppe fungierte das polymergebundene Adamantanmolekül. Mittels Trübungsphotometrie wurden verschiedene Additive auf ihren Einfluss auf den LCST-Phasenübergang der synthetisierten Copolymer untersucht. Hier konnte gezeigt werden, dass die adamantanhaltigen LCST-Copolymere durch Cyclodextrinzusätze in ihrem LCST-Verhalten drastisch zu beinflussen sind. Damit konnte ein System gefunden werden, das mit einfachen analytischen Mitteln eine supramolekulare Erkennung ermöglicht.

Terrylendiimide als Biolabel und funktionelle Farbstoffe

In dieser Arbeit werden neue Rylenimide und Anwendungsmöglichkeiten für diese Farbstoffklasse beschrieben, die sich durch hohe Photostabilitäten und hohe Fluoreszenzquantenausbeute auszeichnet. Ziel dieser Arbeit war es, durch systematische Wahl der Substituenten in den Imidstrukturen und/oder den bay-Regionen von Rylendiimidfarbstoffen vollkommen neue Produkteigenschaften zu verwirklichen, Reaktionen bzw. Anwendungen zu ermöglichen und den Aufbau von komplexeren Chromophorarchitekturen zu gestatten. Das Strukturmotiv des Terrylendiimids nahm dabei die zentrale Rolle ein. Die Arbeit wurde in vier Kapitel aufgeteilt. Das Ziel des ersten Kapitels war es, wasserlösliche Terrylendiimide zur Untersuchung von biologischen Proben im Wellenlängenbereich über 600 nm einzusetzen. Ein wasserlösliches Terrylendiimid erwies sich dabei als deutlich photostabiler als zwei weitverbreitete Fluoreszenzfarbstoffe. Eine erste Proteinmarkierung mit monofunktionellem Farbstoff wurde an Proteinmolekülen erfolgreich durchgeführt. Durch gezielte Modifikationen konnten zwei Terrylendiimide hergestellt werden, die sich noch deutlich besser zum Abbilden von Zellstrukturen eignen. In dem zweiten Kapitel spielte die Löslichkeit von Rylendiimiden in organischen Lösungsmitteln eine zentrale Rolle. Es wurde eine Rylendiimidserie hergestellt, deren löslichkeitssteigernde Gruppen eine Organisation der Moleküle in ausgedehnten Stapelstrukturen nicht verhindern. Mit dieser Serie konnte das flüssigkristalline Verhalten und die Selbstorganisation in der Rylendiimidreihe untersucht werden. Aufbauend auf diesen Ergebnissen wurde die Selbstorganisation der Diimide in Donor-Akzeptor Gemischen untersucht. In STM-Experimenten konnten für alle drei Diimide selbstorganisierte Monoschichten mit dem Rastertunnelmikroskop mit molekularer Auflösung abgebildet werden. Darüber hinaus wurden in diesem Kapitel die ersten organischen Feldeffekttransistoren (OFET) auf der Basis des synthetisierten Terrylendiimids beschrieben. Im Rahmen eines Projektes in dem die elektronische Energieübertragung in Donor-Akzeptor-Diaden mit Hilfe von Einzelmolekülspektroskopie untersucht wird, wurde eine Perylendiimid-Terrylediimid Diade hergestellt. Die geringere Photostabilität des Donors ermöglichte zeitaufgelöste Einzelmolekül-messungen der Akzeptoremission mit und ohne Energietransfer vom Donor auf den Akzeptor. Durch diese Messungen konnten die Zeitkonstanten des Energietransfers für einzelne Diaden ermittelt werden. Ein weiterer Chromophor aus diesem Donor-Akzeptor-Paar soll die Möglichkeit eröffnen, den Energiefluß im Molekül gezielt zu manipulieren. Dazu wurde ein Donorchromophor mit zwei Akzeptoren in einem Multichromophor kombiniert. Im Rahmen der Synthesen dieser Arbeit wurden Terrylendiimide hergestellt, die in einer Imidstruktur eine Halogenfunktion trugen. Diese waren wichtige Synthesebausteine zum Aufbau von komplexen Chromophorarchitekturen. Ziel eines weiteren Kapitels war es, ein Terrylendiimid herzustellen, das als Sensibilisatorfarbstoff gemeinsam mit dem Haupt-Antennenkomplex von höheren Pflanzen LHCII in einer photoelektrochemischen Farbstoff-Solarzelle integriert werden konnte. Das hergestellte Terrylendiimid mit Carbonsäuregruppe eignete sich für Farbstoffsolarzellen auf Zinndioxidbasis.

Strukturbildung durch Selbstorganisation von Oligonukleotiden

Die DNA-Doppelhelix ist eine relativ dicke (Ø ≈ 2 nm), kompakte und dadurch auf kurzen Längenskalen relativ steife Verbindung (lp[dsDNA] ≈ 50-60 nm), mit einer klar definierten Struktur, die durch biologische Methoden sehr präzise manipuliert werden kann. Die Auswirkungen der primären Sequenz auf die dreidimensionale Strukturbildung ist gut verstanden und exakt vorhersagbar. Des Weiteren kann DNA an verschiedenen Stellen mit anderen Molekülen verknüpft werden, ohne dass ihre Selbsterkennung gestört wird. Durch die helikale Struktur besteht außerdem ein Zusammenhang zwischen der Lage und der räumlichen Orientierung von eingeführten Modifikationen. Durch moderne Syntheseverfahren lassen sich beliebige Oligonukleotidsequenzen im Bereich bis etwa 150-200 Basen relativ preiswert im Milligrammmaßstab herstellen. Diese Eigenschaften machen die DNA zu einem idealen Kandidaten zur Erzeugung komplexer Strukturen, die durch Selbsterkennung der entsprechenden Sequenzen gebildet werden. In der hier vorgelegten Arbeit wurden einzelsträngige DNA-Abschnitte (ssDNA) als adressierbare Verknüpfungsstellen eingesetzt, um verschiedene molekulare Bausteine zu diskreten nicht periodischen Strukturen zu verbinden. Als Bausteine dienten flexible synthetische Polymerblöcke und semiflexible Doppelstrang-DNA-Abschnitte (dsDNA), die an beiden Enden mit unterschiedlichen Oligonukleotidsequenzen „funktionalisiert“ sind. Die zur Verknüpfung genutzten Oligonukleotidabschnitte wurden so gewählt (n > 20 Basen), dass ihre Hybridisierung zu einer bei Raumtemperatur stabilen Doppelstrangbildung führt. Durch Kombination der Phosphoramiditsynthese von DNA mit einer festkörpergestützten Blockkopplungsreaktion konnte am Beispiel von Polyethylenoxiden ein sehr effektiver Syntheseweg zur Herstellung von ssDNA1-PEO-ssDNA2-Triblockcopolymeren entwickelt werden, der sich problemlos auf andere Polymere übertragen lassen sollte. Die Längen und Basenabfolgen der beiden Oligonukleotidsequenzen können dabei unabhängig voneinander frei gewählt werden. Somit wurden die Voraussetzungen geschaffen, um die Selbsterkennung von Oligonukleotiden durch Kombination verschiedener Triblockcopolymere zur Erzeugung von Multiblockcopolymeren zu nutzen, die mit klassischen Synthesetechniken nicht zugänglich sind. Semiflexible Strukturelemente lassen sich durch die Synthese von Doppelstrangfragmenten mit langen überstehenden Enden (sticky-ends) realisieren. Die klassischen Ansätze der molekularen Genetik zur Erzeugung von sticky-ends sind in diesem Fall nicht praktikabel, da sie zu Einschränkungen im Bezug auf Länge und Sequenz der überhängenden Enden führen. Als Methode der Wahl haben sich zwei verschiedene Varianten der Polymerase Kettenreaktion (PCR) erwiesen, die auf der Verwendung von teilkomplementären Primern beruhen. Die eigentlichen Primersequenzen wurden am 5´-Ende entweder über ein 2´-Desoxyuridin oder über einen kurzen Polyethylenoxid-Spacer (n = 6) mit einer frei wählbaren „sticky-end-Sequenz“ verknüpft. Mit diesen Methoden sind sowohl 3´- als auch 5´-Überhänge zugänglich und die Länge der Doppelstrangabschnitte kann über einen breiten Molmassenbereich sehr exakt eingestellt werden. Durch Kombination derartiger Doppelstrangfragmente mit den biosynthetischen Triblockcopolymeren lassen sich Strukturen erzeugen, die als Modellsysteme zur Untersuchung verschiedener Biomoleküle genutzt werden können, die in Form eines mehrfach gebrochenen Stäbchens vorliegen. Im letzten Abschnitt wurde gezeigt, dass durch geeignete Wahl der überstehenden Enden bzw. durch Hybridisierung der Doppelstrangfragmente mit passenden Oligonukleotiden verzweigte DNA-Strukturen mit Armlängen von einigen hundert Nanometern zugänglich sind. Im Vergleich zu den bisher veröffentlichten Methoden bietet diese Herangehensweise zwei entscheidende Vorteile: Zum einen konnte der Syntheseaufwand auf ein Minimum reduziert werden, zum anderen ist es auf diesem Weg möglich die Längen der einzelnen Arme, unabhängig voneinander, über einen breiten Molmassenbereich zu variieren.

Modes of Diversification in Bulgarian Higher Education: Case Studies in four Disciplines

This project was stimulated by the unprecedented speed and scope of changes in Bulgarian higher education since 1989. The rapid growth of the student population and the emergence of a new private sector in higher education led to tightening governmental control and a growing criticism of autonomy and academic freedom. This raised questions about the need for diversification in the field, about the importance of recent innovations in terms of strategic choices for future development and so of how higher education governance could maintain diversity without the system deteriorating. The group first traced the extent of spontaneous processes of innovation at the level of content, of institutions, and the organisation of teaching and learning processes. They then identified the different parties in the struggle for institutionalisation and against diversification, and promising mechanisms for maintaining diversity in higher education. On this basis they outlined a basis for a wide-ranging public discussion of the issue which may serve as a corrective to the mechanisms of state control. Their work included analysis of the legislative framework laid down in the Higher Education Act, which effectively dispenses with the autonomy of universities. They then surveyed the views of both high-level executives in the field and the academics actually involved in the process, as well as of the "consumers" of the educational product, i.e. the students. In considering diversification, they focused on four different types of programmes, including those where diversification is largely limited to content level (e.g. Law), those where it operates mainly on structural levels (e.g. Industrial Management), those where it is often feigned (e.g. Social Work), and those where it is at best formal and sporadic (e.g. Mechanical Engineering). They conclude that the educational system in Bulgaria has considerable internal resources for development. The greatest need is for adequate statutory regulation of academic life which will provide incentives for responsible academic development of higher education institutions and create conditions for the institutionalisation of academic self-organisation and self-control, which will in turn limit the pathological trends in the diversification processes.

Compatibility of ultra high performance concrete as repair material : bond characterization with concrete under different loading scenarios

The rehabilitation of concrete structures, especially concrete bridge decks, is a major challenge for transportation agencies in the United States. Often, the most appropriate strategy to preserve or rehabilitate these structures is to provide some form of a protective coating or barrier. These surface treatments have typically been some form of polymer, asphalt, or low-permeability concrete, but the application of UHPC has shown promise for this application mainly due to its negligible permeability, but also as a result of its excellent mechanical properties, self-consolidating nature, rapid gain strength, and minimal creep and shrinkage characteristics. However, for widespread acceptance, durability and performance of the composite system must be fully understood, specifically the bond between UHPC and NSC often used in bridge decks. It is essential that the bond offers enough strength to resist the stress due to mechanical loading or thermal effects, while also maintaining an extended service-life performance. This report attempts to assess the bond strength between UHPC and NSC under different loading configurations. Different variables, such as roughness degree of the concrete substrates, age of bond, exposure to freeze-thaw cycles and wetting conditions of the concrete substrate, were included in this study. The combination of splitting tensile test with 0, 300, 600 and 900 freeze-thaw cycles was carried out to assess the bond performance under severe ambient conditions. The slant-shear test was utilized with different interface angles to provide a wide understanding of the bond performance under different combinations of compression and shear stresses. The pull-off test is the most accepted method to evaluate the bond strength in the field. This test which studies the direct tensile strength of the bond, the most severe loading condition, was used to provide data that can be correlated with the other tests that only can be used in the laboratory. The experimental program showed that the bond performance between UHPC and NSC is successful, as the strength regardless the different degree of roughness of the concrete substrate, the age of the composite specimens, the exposure to freeze-thaw cycles and the different loading configurations, is greater than that of concrete substrate and largely satisfies with ACI 546.3R-06.

Decentralisation and Social Services in England

Although Great Britain is not normally credited with the achievement of having been the first nation state to implement measures characteristic of a welfare state (this honour goes to Germany and Bismarck's strategy of promoting social insurance in the 1880s) it nevertheless pioneered many models of welfare services in view of the early onset of industrialisation in that country and the subsequent social problems it created. Organisations like the Mutual Insurance and Friendly Societies, the Charity Organisation Society or the Settlement Movement characterised an early approach to welfare that is based on initiatives at the civil society level and express a sense of self-help or of self-organisation in such a way that it did not involve the state directly. The state, traditionally, dealt with matters of discipline and public order, and for this reason institutions like prisons and workhouses represented the other end of the scale of 'welfare' provisions.

Nonviolent factors: Variables associated with ethnic violence, and factors which are associated with nonviolent response to grievances toward members of other ethnic groups

Ethnic violence appears to be the major source of violence in the world. Ethnic hostilities are potentially all-pervasive because most countries in the world are multi-ethnic. Public health's focus on violence documents its increasing role in this issue.^ The present study is based on a secondary analysis of a dataset of responses by 272 individuals from four ethnic groups (Anglo, African, Mexican, and Vietnamese Americans) who answered questions regarding variables related to ethnic violence from a general questionnaire which was distributed to ethnically diverse purposive, nonprobability, self-selected groups of individuals in Houston, Texas, in 1993.^ One goal was psychometric: learning about issues in analysis of datasets with modest numbers, comparison of two approaches to dealing with missing observations not missing at random (conducting analysis on two datasets), transformation analysis of continuous variables for logistic regression, and logistic regression diagnostics.^ Regarding the psychometric goal, it was concluded that measurement model analysis was not possible with a relatively small dataset with nonnormal variables, such as Likert-scaled variables; therefore, exploratory factor analysis was used. The two approaches to dealing with missing values resulted in comparable findings. Transformation analysis suggested that the continuous variables were in the correct scale, and diagnostics that the model fit was adequate.^ The substantive portion of the analysis included the testing of four hypotheses. Hypothesis One proposed that attitudes/efficacy regarding alternative approaches to resolving grievances from the general questionnaire represented underlying factors: nonpunitive social norms and strategies for addressing grievances--using the political system, organizing protests, using the system to punish offenders, and personal mediation. Evidence was found to support all but one factor, nonpunitive social norms.^ Hypothesis Two proposed that the factor variables and the other independent variables--jail, grievance, male, young, and membership in a particular ethnic group--were associated with (non)violence. Jail, grievance, and not using the political system to address grievances were associated with a greater likelihood of intergroup violence.^ No evidence was found to support Hypotheses Three and Four, which proposed that grievance and ethnic group membership would interact with other variables (i.e., age, gender, etc.) to produce variant levels of subgroup (non)violence.^ The generalizability of the results of this study are constrained by the purposive self-selected nature of the sample and small sample size (n = 272).^ Suggestions for future research include incorporating other possible variables or factors predictive of intergroup violence in models of the kind tested here, and the development and evaluation of interventions that promote electoral and nonelectoral political participation as means of reducing interethnic conflict. ^