1,354,690 research outputs found
Diseño de modelos fundamentados para la clasificación en problemas con costes dependientes del ejemplo
Get PDFLos problemas de clasificación de patrones pertenecen a los campos del aprendizaje automático y la inteligencia artificial. Cuando se toma una decisión sobre un patrón, generalmente se produce un efecto o consecuencia. Este efecto puede ser positivo (beneficio) o negativo (coste) en función de la calidad de la decisión. La clasificación con costes dependientes de la muestra (EDC, “Example-Dependent Cost”) es un caso particular de clasificación de patrones, en donde estos costes o beneficios varían en función del patrón. En general, los problemas de clasificación EDC suelen estar desequilibrados. Esta tesis doctoral se centrará en el estudio de los problemas de clasificación EDC. Tras realizar una revisión de la literatura, observamos que muchos de los métodos disponibles para diseñar clasificadores EDC no están fundamentados, es decir, no tienen una base teórica sólida, lo cual implica riesgos de degradación. Por otro lado, poder explicar las decisiones de un modelo, es un aspecto fundamental para la implantación de estas técnicas en la sociedad, especialmente en decisiones críticas. Las explicaciones contrafactuales son muy naturales para los humanos. Adicionalmente, son compatibles con modelos complejos como las redes neuronales. No obstante, las técnicas del estado del arte en este ámbito no tienen en cuenta los costes de decisión. La meta de esta tesis doctoral es descubrir las mejores estrategias para maximizar los beneficios durante un proceso de toma de decisiones. Para ello, se pretende proponer una solución completa fundamentada para los problemas de clasificación EDC, la cual incluya tanto métodos para diseñar clasificadores EDC, como métodos para explicar sus decisiones. Para lograrlo, se explorarán tres líneas de investigación: La primera línea es el diseño de clasificadores EDC con costes informados, donde proponemos una nueva metodología Bayesiana en dos pasos que permite estimar de forma fundamentada las probabilidades “a posteriori”. Se exploran diferentes esquemas y reequilibrados neutrales para proponer cuatro nuevos modelos: WR-MLP, WSR-MLPE, WSR-DNN yWSR-DNNE. La segunda línea es el diseño de clasificadores EDC con costes no informados, donde proponemos OsC-MLP, un nuevo método Bayesiano en un paso. El método propone un cociente de verosimilitudes equivalente, y utiliza una función de pérdida que permite a una red neuronal aprender implícitamente los costes de decisión, permitiendo que el umbral no incluya los costes para nuevas muestras. La formulación propuesta incluye esquemas de reequilibrado neutral. La tercera línea es la explicabilidad EDC, donde proponemos COCOA, un nuevo método para generar explicaciones contrafactuales que tengan en cuenta los costes. La novedad principal del método COCOA es guiar la búsqueda con una optimización del discriminante de Bayes. En cada línea de investigación, se realizan experimentos para comparar los beneficios y imitaciones de los métodos propuestos frente a otros métodos de la literatura. Diversas publicaciones [Mediavilla-Relaño et al., 2019, Mediavilla-Relaño et al., 2023, Mediavilla-Relaño y Lázaro, 2024a, Mediavilla-
Relaño y Lázaro, 2024b] avalan las aportaciones de esta tesis doctoral.Pattern recognition problems belong to the fields of machine learning and artificial intelligence. When a decision is made about a pattern, there is usually an effect or consequence. This effect can be positive (benefit) or negative (cost) depending on the quality of the decision. Example-dependent cost (EDC) classification is a particular case of pattern recognition, where these costs or benefits are specific for each pattern. In general, EDC classification problems are usually imbalanced. This doctoral thesis will focus on the study of EDC classification problems. After reviewing the literature, we observed that many methods available to design EDC classifiers are not principled, i.e., they do not have a solid theoretical foundation, which implies risks of degradation. On the other hand, explaining a model’s decisions is a fundamental aspect of implementing these techniques in society, especially in critical decisions. Counterfactual explanations are very natural to humans. Additionally, they are suitable with complex models such as neural networks. However, state-of-the-art techniques in this area do not take decision costs into account. The goal of this doctoral thesis is to discover the best strategies to maximize benefits during a decision-making process. To this end, it is intended to propose a complete well-founded solution for EDC classification problems, which includes both methods to design EDC classifiers and methods to explain their decisions. To achieve this, three lines of research will be explored: The first line is the design of EDC classifiers with informed costs, where we propose a new two-step Bayesian methodology that allows the principled estimation of the “a posteriori” probabilities. Different schemes and neutral rebalancing are explored to propose four new models: WR-MLP, WSR-MLPE, WSR-DNN, and WSR-DNNE. The second line is the design of EDC classifiers with uninformed costs, where we propose OsC-MLP, a new one-step Bayesian method. The method proposes an equivalent likelihood ratio and uses a loss function that allows a neural network to learn the decision costs implicitly, and proposes a test that does not require the costs for new samples. The proposed formulation includes neutral rebalancing schemes. The third line is EDC explainability, where we propose COCOA, a new method for generating cost-aware counterfactual explanations. The main contribution of the COCOA method is to guide the search with an optimization of the Bayes discriminant. In each line of research, experiments are performed to compare the benefits and limitations of the proposed methods against some benchmark methods of the literature. Various publications
[Mediavilla-Relaño et al., 2019, Mediavilla-Relaño et al., 2023, Mediavilla-Relaño y Lázaro, 2024a, Mediavilla-Relaño y Lázaro, 2024b] endorse the contributions of this doctoral thesis.El trabajo de investigación que da lugar al Capítulo 2 y al artículo de revista [Mediavilla-Relaño et al., 2023] ha sido parcialmente financiado por el Proyecto de Investigación 2-BARBAS (Fundación BBVA). El trabajo de investigación que permite la publicación del artículo de revista [Mediavilla-Relaño y Lázaro, 2024b] y al Capítulo 3 ha sido parcialmente apoyado por el Proyecto de Generación de Conocimiento PID2021- 125652OB-I00 (ML-SAM) financiado por la Agencia Estatal de Investigación MCIN/AEI/10.13039/501100011033 y por “ERDF A way of making Europe”. Adicionalmente los costes de publicación en abierto (APC, “Article Processing Charge”) han sido financiados por la Universidad Carlos III de Madrid a través de los acuerdos transformativos establecidos entre la editorial y la Conferencia de Rectores de Universidades Españolas (CRUE) y el Consorcio Madroño (Acuerdo CRUE-Madroño 2024).Programa de Doctorado en Multimedia y Comunicaciones por la Universidad Carlos III de Madrid y la Universidad Rey Juan CarlosPresidente: Francisco Herrera Triguero.- Secretario: Isaac Martín de Diego.- Vocal: Gonzalo Martínez Muño
Recognizing religious minority rights: a key tool to managing pluralism in the European Court of Human Rights
No full textThe article examines how the European Court of Human Rights (ECtHR) engages with the rights of religious minorities within the framework of the European Convention on Human Rights (ECHR). Despite the absence of explicit provisions for minority rights in the ECHR, the Court has navigated this gap by leveraging the principles of pluralism and diversity. The analysis reveals a complex interplay between individual and group rights, highlighting both the progress and limitations in the ECtHR's jurisprudence.
The author introduces a theoretical framework to categorize religious minority rights, emphasizing the intrinsic and collective dimensions of these rights. The article critiques the ECtHR for inconsistent rulings, particularly when religious expressions conflict with state interests such as secularism or public order. Furthermore, it underscores the challenges of balancing individual autonomy with the collective identity of religious groups. By dissecting key rulings, the article exposes the Court's struggles with protecting "uncomfortable diversities," such as Islamic practices, while upholding democratic values. The article concludes by advocating for a more empathetic and pluralistic approach in adjudicating religious minority rights. It calls for the ECtHR to transcend stereotypes and abstract principles, fostering a legal landscape that values diversity and inclusivity as pillars of a democratic society. This shift, the author argues, is essential for respecting the rights and dignity of all, and for achieving a genuine coexistence of plural identities in Europe.Depto. de Derecho Internacional, Eclesiástico y Filosofía del DerechoFac. de DerechoTRUEpu
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Get PDF“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Get PDFWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
Is multiculturalism a satisfactory framework to address religious diversity?
No full textThe chapter, "Is Multiculturalism a Satisfactory Framework to Address Religious Diversity?", delves into the challenges and limitations of traditional multiculturalism in managing religious diversity within contemporary pluralistic societies. It critiques the inadequacies of the multiculturalist model in addressing the complexities of evolving cultural and religious identities, particularly amidst rising intolerance, Islamophobia, and anti-Semitism. The chapter highlights the tension between majority cultural dominance and the need to safeguard minority religious identities, emphasizing the importance of legal recognition and inclusivity.
The discussion explores how globalization, migration, and secularization have intensified religious diversity, making the management of "everyday religious superdiversity" a pressing necessity. It critiques liberal secularism for its "blindness to difference," often marginalizing minority religious practices under the guise of cultural neutrality. Furthermore, it examines the rise of alternative paradigms like interculturalism and postmulticulturalism, which aim to bridge the gaps left by traditional models by promoting dialogue, shared values, and the coexistence of individual and collective identities.
Central to the chapter is the concept of human dignity, which it positions as the foundation for addressing religious diversity. This dignity-based framework asserts that individual and collective identities must be recognized equally, and that public policies must reflect this ethos. The proposed "everyday religious superdiversity" framework combines structural (top-down) approaches, such as implementing international human rights standards, with individual (bottom-up) approaches that value daily religious practices and personal autonomy.
The chapter concludes with a call for reinterpreting traditional notions like secularism and neutrality to accommodate the realities of religious pluralism in public spaces. By integrating inclusive and diversity-sensitive policies, it outlines a vision for a more equitable, cohesive, and pluralistic society, capable of addressing the challenges of modern religious diversity.Depto. de Derecho Internacional, Eclesiástico y Filosofía del DerechoFac. de DerechoTRUEpu
Dispelling the Myths Behind First-author Citation Counts
Get PDFWe conducted a full-scale evaluative citation analysis study of scholars in the XML research field to explore just how different from each other author rankings resulting from different citation counting methods actually are, and to demonstrate the capability of emerging data and tools on the Web in supporting more realistic citation counting methods. Our results contest some common arguments for the continued
use of first-author citation counts in the evaluation of scholars, such as high correlations between author rankings by first-author citation counts and other citation
counting methods, and high costs of using more realistic citation counting methods that are not well-supported by the ISI databases. It is argued that increasingly available digital full text research papers make it possible for citation analysis studies to go beyond what the ISI databases have directly supported and to employ more
sophisticated methods
Supplementary material for: Evaluating the role of age on speech-in-noise perception based primarily on temporal envelope information
No full textThis project holds the supplementary material related to the Psyarxiv preprint: Regev, J.; Oxenham, A. J.; Relaño-Iborra, H.; Zaar, J.; and Dau, T. "Evaluating the role of age on speech-in-noise perception based primarily on temporal envelope information" https://doi.org/10.31234/osf.io/uf5x
Author, publisher and bookseller : a tripartite synergy in Nigerian book industry
Get PDFThis work is about the roles of Author, Publisher and Bookseller in Book development in
Nigeria. The paper started by delving into the history of Book Publishing in Nigeria after
which it proceeded by defining who an author, a publisher, and a bookseller is and
expatiated on the indispensable roles of these key actors in Nigerian Book Industry and in
the emerging Information Society. Furthermore, the various constraints to book
development were identified while the paper advised on how the Book Industry can be
further promoted in Nigeria. However, the paper concluded and made recommendations
on how the Book sector can help in enhancing scholarship in the country
The Thursday Murder Club: Launching a megabrand author - a publishing case study
Get PDFIn 2020, the Christmas book charts in the UK made headlines: Barack Obama’s eagerly awaited autobiography, The Promised Land, was beaten to the top spot by The Thursday Murder Club by Richard Osman, a debut cosy crime novel set in a retirement village. Not only did Osman’s book beat the former US president’s expected bestseller, it also broke records, becoming the fastest-selling debut crime novel of all time. Although Osman has a certain level of fame in the UK from his TV appearances on shows such as Pointless, his celebrity status does not entirely explain the novel’s huge sales. This article tracks the acquisition, publication, and promotion journey of The Thursday Murder Club in order to understand the industry and cultural context of its success and to interrogate the role of celebrity in the creation of author brands. The findings suggest that the unexpected scale of the success of the book owed to a number of factors, including in-depth editing by the novel’s agent, editor, and author to tighten up the plot, an extensive and strategic promotional campaign, the pandemic (which drove interest in the book’s genre and themes), and the quality of the writing. We find that the book’s success was accentuated by Osman’s celebrity status rather than being entirely reliant on it. This research adds to the growing scholarship on celebrity authorship by means of an in-depth case study and provides insight into the processes behind publishing a ‘celebrity’ book and launching a megabrand author
- …
