1,720,962 research outputs found
Método adaptativo en tiempo real para la detección de anomalías mediante aprendizaje automático
Get PDF[Resumen] La detección de anomalías es la subrama del aprendizaje automático encargada de construir modelos capaces de diferenciar entre datos normales y anómalos. Ya que los datos normales son los que prevalecen en estos escenarios y sus características suelen ser bien conocidas, el entrenamiento de los sistemas se realiza mayormente mediante estos conjuntos de datos normales, lo que diferencia a la detección de anomalías de otros problemas de clasificación estándar. Debido al habitual uso de estos sistemas en monitorización y a la inexistencia de métodos capaces de aprender en tiempo real, en este proyecto de investigación se presenta un nuevo método que proporciona dicha capacidad de adaptación online. El método desarrollado recibe el nombre de OnlineS-DSCH (Online and Subdivisible Distributed Scaled Convex Hull) y basa su funcionamiento en las propiedades de los cierres convexos. Tras su desarrollo, se ha evaluado y comparado su rendimiento con los principales algoritmos del área sobre diferentes conjuntos de datos, tanto reales como artificiales. Como consecuencia, se ha obtenido un algoritmo con la capacidad de aprendizaje online, fácilmente configurable y cuyas predicciones son justificables, todo ello sin que suponga una merma en su eficacia en relación a las otras soluciones disponibles. Por último, su ejecución se puede llevar a cabo de manera distribuida y en paralelo, lo que supone una ventaja interesante en el tratamiento de conjuntos de datos de alta dimensionalidad.[Abstract] In machine learning, anomaly detection is the branch responsible for building models capable of differentiating between normal and anomalous data. Normal data prevail in scenarios of this type, and their features are usually well known, so the training phase is largely done through these normal data sets, which differentiates the anomaly detection from other standard classification problems. Due to the usual use of these systems in monitoring and the lack of methods capable of learning in real time, this research project presents a new method that provides such online adaptability. The method developed is called OnlineS-DSCH (Online and Subdivisible Distributed Scaled Convex Hull) and bases its operation on the properties of convex hulls. After its development, its performance has been evaluated and compared with the main algorithms of the area on different real and artificial data sets. As a consequence, an algorithm with the online learning capacity, easily configurable and whose predictions are justifiable, has been obtained, all without diminishing its effectiveness in relation to the other available solutions. Finally, its execution can be carried out in a distributed and parallel way, which is an interesting advantage in the treatment of high dimensionality data sets.Traballo fin de grao (UDC.FIC). Enxeñaría informática. Curso 2018/201
Machine Learning for Anomaly Detection: From Surface to Deep
Get PDF[Resumo] A detección de anomalías é a rama da aprendizaxe automática encargada de construír modelos capaces de diferenciar entre datos normais e anómalos. A priori,
isto converte a detección de anomalías nun problema de clasificación en dúas clases.
Con todo, dado que as anomalías adoitan ocorrer de forma esporádica, os datos
normais son os que prevalecen nestes escenarios, polo que é habitual que se requiran
modelos específicos cuxo adestramento sexa levado a cabo empregando unicamente
datos da clase normal. Estes sistemas xogan un papel vital nunha ampla gama de
aplicacións reais, como a medicina, a detección de fraudes bancarias, intrusións na
rede, ou o mantemento preditivo de sistemas industriais.
Esta tese centrouse no desenvolvemento de novos algoritmos de detección de
anomalías para tres escenarios diferentes, desde modelos baseados na xeometría dos
datos, ata modelos máis complexos baseados en aprendizaxe profunda. No primeiro
escenario propoñemos un método baseado en peches non convexos subdivisibles
para escenarios tradicionais, nos que algúns dos principais problemas son a dimensionalidade
dos datos de adestramento e a forma da nube de puntos no espacio
n-dimensional. No segundo escenario preséntase unha rede autoencoder profunda
válida para escenarios de computación no borde e aprendizaxe federado debido ao
seu adestramento non iterativo, así como unha arquitectura federada para a súa
implementación semi-centralizada. Por último, en liña coas novas tendencias da intelixencia
artificial aplicable, estudouse a aplicabilidade da detección de anomalías
sobre textos, propoñendo un pipeline para a detección de recensións anómalas explicable
en plataformas de comercio electrónico.
O traballo desenvolvido foi compartido coa comunidade investigadora a través de
publicacións en revistas e conferencias científicas, ademais de repositorios de código
aberto, contribuíndo así ao avance do campo da detección de anomalías.[Resumen] La detección de anomalías es la rama del aprendizaje automático encargada
de construir modelos capaces de diferenciar entre datos normales y anómalos. A
priori, esto convierte la detección de anomalías en un problema de clasificación en
dos únicas clases. Sin embargo, dado que las anomalías suelen ocurrir de forma
esporádica, los datos normales son los que prevalecen en estos escenarios, por lo que
es habitual que se requieran modelos específicos cuyo entrenamiento se lleve a cabo
empleando únicamente datos de la clase normal. Estos sistemas juegan un papel
vital en una amplia gama de aplicaciones reales, como la medicina, la detección de
fraudes bancarios, intrusiones en la red, o el mantenimiento predictivo de sistemas
industriales.
Esta tesis se ha centrado en el desarrollo de nuevos algoritmos de detección de
anomalías para tres escenarios diferentes, desde modelos basados en la geometría
de los datos, hasta modelos más complejos basados en aprendizaje profundo. En el
primer escenario proponemos un método basado en cierres no convexos subdivisibles
para entornos tradicionales, en los que algunos de los principales problemas son la
dimensionalidad de los datos de entrenamiento y la forma de la nube de puntos en
el espacio n-dimensional. En el segundo escenario se presenta una red autoencoder
profunda válida para entornos de computación en el borde y aprendizaje federado
debido a su entrenamiento no iterativo, así como una arquitectura federada para su
implementación semi-centralizada. Por último, en línea con las nuevas tendencias
de inteligencia artificial aplicable, se ha estudiado la aplicabilidad de la detección
de anomalías sobre textos, proponiendo un pipeline para la detección de reseñas
anómalas explicable en plataformas de comercio electrónico.
El trabajo desarrollado ha sido compartido con la comunidad investigadora a
través de publicaciones en revistas y congresos científicos, así como repositorios de
código abierto, contribuyendo de este modo al avance del campo de la detección de
anomalías.[Abstract] Anomaly detection is the branch of machine learning in charge of building models capable of differentiating between normal and anomalous data. A priori, this makes anomaly detection a two-class classification problem. However, since anomalies tend to occur sporadically, normal data are the ones that prevail in these scenarios, so
it is common to require specific models whose training is carried out using only
data from the normal class. These systems play a vital role in a wide range of realworld
applications, such as medicine, bank fraud detection, network intrusions, or
predictive maintenance of industrial systems.
This thesis has focused on the development of new anomaly detection algorithms
for three different scenarios, from models based on data geometry to more
complex models based on deep learning. In the first scenario, we propose a method
based on subdivisible non-convex hulls for traditional environments, where some
of the main problems are the dimensionality of the training data and the shape
of the point cloud in the n-dimensional space. In the second scenario, a deep autoencoder
network valid for edge computing and federated learning environments is presented due to its non-iterative training, as well as a federated architecture for
its semi-centralized implementation. Finally, in accordance with the new trends in
applied artificial intelligence, the applicability of anomaly detection on text has been
studied, proposing a pipeline for the detection of explainable anomalous reviews in
e-commerce platforms.
The work developed has been shared with the research community through publications in scientific journals and conferences, as well as open source repositories,
thus contributing to the advancement of the field of anomaly detection
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
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
koamabayili/VECTRON-author-checklist: VECTRON author checklist
No full textWe have done our best to complete the author checklist relating to the use of animals in the hut study. Note that the objective for the hut study was to evaluate the IRS treatment applications for residual efficacy against Anopheles mosquitoes, including the local An. coluzzii mosquito population. Cows were only used to attract mosquitoes into the huts and no tests were carried out directly on the cows. The author checklist is intended for use with studies where experiments are carried out on animals, which is why we have had such difficulty in completing this for the hut study, as many of the questions do not relate to how the cows were used
Author-wise bibliometric analysis based on entropy.
No full textAuthor-wise bibliometric analysis based on entropy.</p
- …
