Search

[Abstract]: Recommender Systems have become crucial in the modern world, commonly guiding users towards relevant content or products, and having a large influence over the decisions of users and citizens. However, ensuring transparency and user trust in these systems remains a challenge; personalized explanations have emerged as a solution, offering justifications for recommendations. Among the existing approaches for generating personalized explanations, using existing visual content created by users is a promising option to maximize transparency and user trust. State-of-the-art models that follow this approach, despite leveraging highly optimized architectures, employ surrogate learning tasks that do not efficiently model the objective of ranking images as explanations for a given recommendation; this leads to a suboptimal training process with high computational costs that may not be reduced without affecting model performance. This work presents BRIE, a novel model where we leverage Bayesian Pairwise Ranking to enhance the training process, allowing us to consistently outperform state-of-the-art models in six real-world datasets while reducing its model size by up to 64 times and its CO2 emissions by up to 75% in training and inference.

[Abstract]: In classical survival analysis, it is assumed that all the individuals will experience the event of interest. However, if there is a proportion of subjects who will never experience the event, then a standard survival approach is not appropriate, and cure models should be considered instead. This paper deals with the problem of adapting a machine learning approach for classical survival analysis to a situation when cure (i.e., not suffering the event) is a possibility. Specifically, a brief review of cure models and recent machine learning methodologies is presented, and an adaptation of machine learning approaches to account for cured individuals is introduced. In order to validate the proposed methods, we present an extensive simulation study in which we compare the performance of the adapted machine learning algorithms with existing cure models. The results show the good behavior of the semiparametric or the nonparametric approaches, depending on the simulated scenario. The practical utility of the methodology is showcased through two real-world dataset illustrations. In the first one, the results show the gain of using the nonparametric mixture cure model approach. In the second example, the results show the poor performance of some machine learning methods for small sample sizes.

[Abstract]: Classic embedded feature selection algorithms are often divided in two large groups: tree-based algorithms and LASSO variants. Both approaches are focused in different aspects: while the tree-based algorithms provide a clear explanation about which variables are being used to trigger a certain output, LASSO-like approaches sacrifice a detailed explanation in favor of increasing its accuracy. In this paper, we present a novel embedded feature selection algorithm, called End-to-End Feature Selection (E2E-FS), that aims to provide both accuracy and explainability in a clever way. Despite having non-convex regularization terms, our algorithm, similar to the LASSO approach, is solved with gradient descent techniques, introducing some restrictions that force the model to specifically select a maximum number of features that are going to be used subsequently by the classifier. Although these are hard restrictions, the experimental results obtained show that this algorithm can be used with any learning model that is trained using a gradient descent algorithm.

[Abstract] In this study, we analyze the capability of several state of the art machine learning methods to predict whether patients diagnosed with CoVid-19 (CoronaVirus disease 2019) will need different levels of hospital care assistance (regular hospital admission or intensive care unit admission), during the course of their illness, using only demographic and clinical data. For this research, a data set of 10,454 patients from 14 hospitals in Galicia (Spain) was used. Each patient is characterized by 833 variables, two of which are age and gender and the other are records of diseases or conditions in their medical history. In addition, for each patient, his/her history of hospital or intensive care unit (ICU) admissions due to CoVid-19 is available. This clinical history will serve to label each patient and thus being able to assess the predictions of the model. Our aim is to identify which model delivers the best accuracies for both hospital and ICU admissions only using demographic variables and some structured clinical data, as well as identifying which of those are more relevant in both cases. The results obtained in the experimental study show that the best models are those based on oversampling as a preprocessing phase to balance the distribution of classes. Using these models and all the available features, we achieved an area under the curve (AUC) of 76.1% and 80.4% for predicting the need of hospital and ICU admissions, respectively. Furthermore, feature selection and oversampling techniques were applied and it has been experimentally verified that the relevant variables for the classification are age and gender, since only using these two features the performance of the models is not degraded for the two mentioned prediction problems.

[Abstract]: This paper presents a unified propagation method for dealing with both the classic Eikonal equation, where the motion direction does not affect the propagation, and the more general static Hamilton-Jacobi equations, where it does. While classic Fast Marching Method (FMM) techniques achieve the solution to the Eikonal equation with a O(M log M) (or O(M) assuming some modifications), solving the more general static Hamilton-Jacobi equation requires a higher complexity. The proposed framework maintains the O(M log M) complexity for both problems, while achieving higher accuracy than available state-of-the-art. The key idea behind the proposed method is the creation of ‘mini wave-fronts’, where the solution is interpolated to minimize the discretization error. Experimental results show how our algorithm can outperform the state-of-the-art both in precision and computational cost.

In this work, we explore in detail and propose a system to cluster faces from unconstrained images. This system can be divided mainly in two big steps: i) align the faces and pass them through a deep convolutional neural network, and ii) cluster the face images by their feature representation.

[Resumen] El análisis de comportamiento humano es uno de los campos más activos en la rama de visión por computador. Con el incremento de cámaras, especialmente en entornos controlados tales como aeropuertos, estaciones de tren o museos, se hace cada vez más necesario el uso de sistemas automáticos que puedan catalogar la información proporcionada. En el caso de entornos concurridos, es muy difícil el poder distinguir el comportamiento de personas en base a sus gestos, debido a la falta de visión de su cuerpo al completo. Por ende, el análisis de comportamiento se realiza en base a sus trayectorias, añadiendo técnicas de razonamiento de alto nivel para ulilizar dicha información en múltiples aplicaciones, tales como la video vigilancia o el análisis de tráfico. El propósito de esta investigación es el desarrollo de un sistema totalmente automático para el análisis de comportamiento de las personas. Por una parte, se presentan dos sistemas para el seguimiento de múltiples objetivos, así como un sistema novedoso para la re-identificación de personas, con la intención de detectar todo objeto de interés en la escena, devolviendo sus trayectorias como salida. Por otra parte, se presenta un sistema novedoso para el análisis de comportamiento basado en información del entorno de la escena. Está basado en la idea que que toda persona,cuando intenta llegar a un cierto lugar, tiende a seguir el mismo camino que suele utilizar la mayoría de la gente. Se presentan una serie de métricas para la detección de movimientos anómalos, haciendo que este método sea ideal para su utilización en sistemas de tiempo real., [Abstract] Human behavior analysis is one of the most active computer vision research fields. As the number of cameras are increased, especially in restricted environments, like airports, train stations or museums, the need of automatic systems that can catalog the information provided by the cameras becomes crucial. In the case of crowded scenes, it is very difficult to distinguish people behavior because of the lack of visual contact of the whole body. Thus, behavior analysis remains in the evaluation of trajectories, adding high-level knowledge approaches in order to use that information in several applications like video surveillance or traffic analysis. The proposal of this research is the design of a fully-automatic human behavior system from a distance. On the one hand, two different multiple-target tracking methods and a target re-identification procedure are presented to detect every target in the scene, returning their trajectories as output. On the other hand, a novel behavior analysis system, which includes information about the environment, is provided. It is based in the idea that every person tries to reach a goal in the scene following the same path the majority of people should use. An extremely fast abnormal behavior metric is presented, providing our method with the capabilities needed to be used in real-time scenarios, [Resumo] A análise do comportamento humano é un dos campos máis activos na rama da visión por computadora. Co incremento de cámaras, especialmente en entornos controlados tales coma aeroportos, estacións de tren ou museos, faise cada vez máis necesario o uso de sistemas automáticos que poidan catalogar a información proporcionada. No caso de entornos concurridos, é moi complicado de poder distinguir o comportamento de persoas dacordo cos seus xestos, debido á falta dunha visión completa do corpo do suxeito. Por tanto, a análise de comportamento tende a realizarse en base á traxectoria, engadindo técnicas de razoamento de alto nivel para utilizar dita información en diversas aplicacións, tales coma a video vixiancia ou a análise de tráfico. O propósito desta investigación é o desenrolo dun sistema totalmente automático para a análise do comportamento das persoas. Por unha parte, preséntanse dous sistemas para o seguimento de múltiples obxectivos, así coma un sistema novidoso para a re-identificación de persoas, coa intención de detectar todo obxecto de interés na escena, devolvendo as traxectorias asociadas como saída. Por outra parte, preséntase un sistema novidoso para a análise de comportamente baseada na informaci ón do entorno da escena. Está baseado na idea de que toda persoa, cando intenta acadar un certo luegar, tende a seguir o mesmo cami~no que xeralmente usa a maioría da xente. Preséntanse unha serie de métricas para a detección de movementos anómalos, facendo posible que este método poida ser utilizado en sistemas de tempo real.

[Resumen] El análisis de comportamiento humano es uno de los campos más activos en la rama de visión por computador. Con el incremento de cámaras, especialmente en entornos controlados tales como aeropuertos, estaciones de tren o museos, se hace cada vez más necesario el uso de sistemas automáticos que puedan catalogar la información proporcionada. En el caso de entornos concurridos, es muy difícil el poder distinguir el comportamiento de personas en base a sus gestos, debido a la falta de visión de su cuerpo al completo. Por ende, el análisis de comportamiento se realiza en base a sus trayectorias, añadiendo técnicas de razonamiento de alto nivel para ulilizar dicha información en múltiples aplicaciones, tales como la video vigilancia o el análisis de tráfico. El propósito de esta investigación es el desarrollo de un sistema totalmente automático para el análisis de comportamiento de las personas. Por una parte, se presentan dos sistemas para el seguimiento de múltiples objetivos, así como un sistema novedoso para la re-identificación de personas, con la intención de detectar todo objeto de interés en la escena, devolviendo sus trayectorias como salida. Por otra parte, se presenta un sistema novedoso para el análisis de comportamiento basado en información del entorno de la escena. Está basado en la idea que que toda persona,cuando intenta llegar a un cierto lugar, tiende a seguir el mismo camino que suele utilizar la mayoría de la gente. Se presentan una serie de métricas para la detección de movimientos anómalos, haciendo que este método sea ideal para su utilización en sistemas de tiempo real., [Abstract] Human behavior analysis is one of the most active computer vision research fields. As the number of cameras are increased, especially in restricted environments, like airports, train stations or museums, the need of automatic systems that can catalog the information provided by the cameras becomes crucial. In the case of crowded scenes, it is very difficult to distinguish people behavior because of the lack of visual contact of the whole body. Thus, behavior analysis remains in the evaluation of trajectories, adding high-level knowledge approaches in order to use that information in several applications like video surveillance or traffic analysis. The proposal of this research is the design of a fully-automatic human behavior system from a distance. On the one hand, two different multiple-target tracking methods and a target re-identification procedure are presented to detect every target in the scene, returning their trajectories as output. On the other hand, a novel behavior analysis system, which includes information about the environment, is provided. It is based in the idea that every person tries to reach a goal in the scene following the same path the majority of people should use. An extremely fast abnormal behavior metric is presented, providing our method with the capabilities needed to be used in real-time scenarios, [Resumo] A análise do comportamento humano é un dos campos máis activos na rama da visión por computadora. Co incremento de cámaras, especialmente en entornos controlados tales coma aeroportos, estacións de tren ou museos, faise cada vez máis necesario o uso de sistemas automáticos que poidan catalogar a información proporcionada. No caso de entornos concurridos, é moi complicado de poder distinguir o comportamento de persoas dacordo cos seus xestos, debido á falta dunha visión completa do corpo do suxeito. Por tanto, a análise de comportamento tende a realizarse en base á traxectoria, engadindo técnicas de razoamento de alto nivel para utilizar dita información en diversas aplicacións, tales coma a video vixiancia ou a análise de tráfico. O propósito desta investigación é o desenrolo dun sistema totalmente automático para a análise do comportamento das persoas. Por unha parte, preséntanse dous sistemas para o seguimento de múltiples obxectivos, así coma un sistema novidoso para a re-identificación de persoas, coa intención de detectar todo obxecto de interés na escena, devolvendo as traxectorias asociadas como saída. Por outra parte, preséntase un sistema novidoso para a análise de comportamente baseada na informaci ón do entorno da escena. Está baseado na idea de que toda persoa, cando intenta acadar un certo luegar, tende a seguir o mesmo cami~no que xeralmente usa a maioría da xente. Preséntanse unha serie de métricas para a detección de movementos anómalos, facendo posible que este método poida ser utilizado en sistemas de tempo real.