Search

Trustworthy Artificial Intelligence (AI) is based on seven technical requirements sustained over three main pillars that should be met throughout the system’s entire life cycle: it should be (1) lawful, (2) ethical, and (3) robust, both from a technical and a social perspective. However, attaining truly trustworthy AI concerns a wider vision that comprises the trustworthiness of all processes and actors that are part of the system’s life cycle, and considers previous aspects from different lenses. A more holistic vision contemplates four essential axes: the global principles for ethical use and development of AI-based systems, a philosophical take on AI ethics, a risk-based approach to AI regulation, and the mentioned pillars and requirements. The seven requirements (human agency and oversight; robustness and safety; privacy and data governance; transparency; diversity, non-discrimination and fairness; societal and environmental wellbeing; and accountability) are analyzed from a triple perspective: What each requirement for trustworthy AI is, Why it is needed, and How each requirement can be implemented in practice. On the other hand, a practical approach to implement trustworthy AI systems allows defining the concept of responsibility of AI-based systems facing the law, through a given auditing process. Therefore, a responsible AI system is the resulting notion we introduce in this work, and a concept of utmost necessity that can be realized through auditing processes, subject to the challenges posed by the use of regulatory sandboxes. Our multidisciplinary vision of trustworthy AI culminates in a debate on the diverging views published lately about the future of AI. Our reflections in this matter conclude that regulation is a key for reaching a consensus among these views, and that trustworthy and responsible AI systems will be crucial for the present and future of our society.

In this paper, three customised Artificial Intelligence (AI) frameworks, considering Deep Learning, Machine Learning (ML) algorithms and data reduction techniques, are proposed for a problem of long-term summer air temperature prediction. Specifically, the prediction of the average air temperature in the first and second August fortnights, using input data from previous months, at two different locations (Paris, France) and (Córdoba, Spain), is considered. The target variable, mainly in the first August fortnight, can contain signals of extreme events such as heatwaves, like the heatwave of 2003, which affected France and the Iberian Peninsula. Three different computational frameworks for air temperature prediction are proposed: a Convolutional Neural Network (CNN), with video-to-image translation, several ML approaches including Lasso regression, Decision Trees and Random Forest, and finally a CNN with pre-processing step using Recurrence Plots, which convert time series into images. Using these frameworks, a very good prediction skill has been obtained in both Paris and Córdoba regions, showing that the proposed approaches can be an excellent option for seasonal climate prediction problems.

Over the last few years, convolutional neural networks (CNNs) have dominated the field of computer vision thanks to their ability to extract features and their outstanding performance in classification problems, for example in the automatic analysis of X-rays. Unfortunately, these neural networks are considered black-box algorithms, i.e. it is impossible to understand how the algorithm has achieved the final result. To apply these algorithms in different fields and test how the methodology works, we need to use eXplainable AI techniques. Most of the work in the medical field focuses on binary or multiclass classification problems. However, in many real-life situations, such as chest X-rays, radiological signs of different diseases can appear at the same time. This gives rise to what is known as ”multilabel classification problems”. A disadvantage of these tasks is class imbalance, i.e. different labels do not have the same number of samples. The main contribution of this paper is a Deep Learning methodology for imbalanced, multilabel chest X-ray datasets. It establishes a baseline for the currently underutilised PadChest dataset and a new eXplainable AI technique based on heatmaps. This technique also includes probabilities and inter-model matching. The results of our system are promising, especially considering the number of labels used. Furthermore, the heatmaps match the expected areas, i.e. they mark the areas that an expert would use to make a decision.

Las matrices origen-destino son una herramienta fundamental dentro del ámbito de la movilidad urbana, ya que describen los patrones de desplazamiento de peatones, transporte público o vehículos a lo largo de un área geográfica. Contar con dicha información permite tomar decisiones más acertadas a la hora de realizar el dimensionamiento de infraestructuras de transporte, la planificación del transporte público o la planificación urbana. Dada su importancia, se han realizado una gran cantidad de estudios acerca de posibles métodos de inferencia en los últimos 50 años. Los primeros planteamientos del problema, debido a la naturaleza de los datos, se encontraban ante un problema altamente subdeterminado, lo cual exigía de la incorporación de información adicional al problema. Gran parte de estos estudios eran únicamente teóricos, o se implementaban en escenarios de tamaño muy reducido. Con el paso de los años, y la aparición de nuevos tipos de datos, la variedad de métodos de inferencia se ha ampliado considerablemente, así como su aplicabilidad a escenarios reales. Sin embargo, estos nuevos tipos de datos también tienen sus inconvenientes y el acceso a los mismos es limitado, por lo que la estimación de matrices origen-destino sigue siendo un problema abierto para muchas empresas del ámbito de la movilidad. En este proyecto, en colaboración con una empresa del ámbito de la movilidad, se teoriza e implementa un método de estimación novedoso a partir de datos convencionales, y se evalúa su funcionamiento en un escenario real.

Electrocardiogram (ECG) is a widely used technique to diagnose cardiovascular diseases. It is a non-invasive technique that represents the cyclic contraction and relaxation of heart muscles. ECG can be used to detect abnormal heart motions, heart attacks, heart diseases, or enlarged hearts by measuring the heart’s electrical activity. Over the past few years, various works have been done in the field of studying and analyzing the ECG signals to detect heart diseases. In this work, we propose a deep learning and fuzzy clustering (Fuzz-ClustNet) based approach for Arrhythmia detection from ECG signals. We started by denoising the collected ECG signals to remove errors like baseline drift, power line interference, motion noise, etc. The denoised ECG signals are then segmented to have an increased focus on the ECG signals. We then perform data augmentation on the segmented images to counter the effects of the class imbalance. The augmented images are then passed through a CNN feature extractor. The extracted features are then passed to a fuzzy clustering algorithm to classify the ECG signals for their respective cardio diseases. We ran intensive simulations on two benchmarked datasets and evaluated various performance metrics. The performance of our proposed algorithm was compared with several recently proposed algorithms for heart disease detection from ECG signals. The obtained results demonstrate the efficacy of our proposed approach as compared to other contemporary algorithms.

Panchromatic and multispectral image fusion, termed pan-sharpening, is to merge the spatial and spectral information of the source images into a fused one, which has a higher spatial and spectral resolution and is more reliable for downstream tasks compared with any of the source images. It has been widely applied to image interpretation and pre-processing of various applications. A large number of methods have been proposed to achieve better fusion results by considering the spatial and spectral relationships among panchromatic and multispectral images. In recent years, the fast development of artificial intelligence (AI) and deep learning (DL) has significantly enhanced the development of pan-sharpening techniques. However, this field lacks a comprehensive overview of recent advances boosted by the rise of AI and DL. This paper provides a comprehensive review of a variety of pan-sharpening methods that adopt four different paradigms, i.e., component substitution, multiresolution analysis, degradation model, and deep neural networks. As an important aspect of pan-sharpening, the evaluation of the fused image is also outlined to present various assessment methods in terms of reduced-resolution and full-resolution quality measurement. Then, we conclude this paper by discussing the existing limitations, difficulties, and challenges of pan-sharpening techniques, datasets, and quality assessment. In addition, the survey summarizes the development trends in these areas, which provide useful methodological practices for researchers and professionals. Finally, the developments in pan-sharpening are summarized in the conclusion part. The aim of the survey is to serve as a referential starting point for newcomers and a common point of agreement around the research directions to be followed in this exciting area.

167 p., Driven by the quest to create intelligent systems that can autonomously learn to make optimal decisions, Reinforcement Learning has emerged as a powerful branch of Machine Learning. Reinforcement Learning agents interact with their environment, learning from trial and error, guided by feedback signals shaped in the form of rewards. However, the application of Reinforcement Learning is often hampered by the complexity associated with the design of such rewards. Creating a dense reward function, where the agent receives immediate and frequent feedback from its actions, is often a challenging task. This challenge arises from the difficulty of specifying the correct behavior for every possible state-action pair. This issue parallels the challenges faced in human learning where educators often grapple with identifying the best way to teach a certain skill or subject, given that learning styles can vary dramatically among individuals. As a consequence, it is common to formulate the problems with sparse rewards, where the agent is only rewarded when it accomplishes a significant task or achieves the final goal, thus aligning more directly with the objective of the problem. The sparse reward formulation does not require the anticipation of every possible scenario or state, making it more tractable for complex environments and real-world scenarios, where feedback is often delayed and not immediately available.However, sparse reward settings also introduce their own challenges, most notably, the issue of exploration. In the absence of frequent rewards, an agent can struggle to identify beneficial actions, making learning slow and inefficient. This is where mechanisms such as Intrinsic Motivation come into play, encouraging more effective exploration and improving sample efficiency, despite the sparsity of extrinsic rewards.In this context, the overall contribution of this Thesis is to delve into how Intrinsic Motivation can boost the performance of Deep Reinforcement Learning

284 p., Artificial Intelligence is a non-stopping field of research that has experienced some incredible growth lastdecades. Some of the reasons for this apparently exponential growth are the improvements incomputational power, sensing capabilities and data storage which results in a huge increment on dataavailability. However, this growth has been mostly led by a performance-based mindset that has pushedmodels towards a black-box nature. The performance prowess of these methods along with the risingdemand for their implementation has triggered the birth of a new research field. Explainable ArtificialIntelligence. As any new field, XAI falls short in cohesiveness. Added the consequences of dealing withconcepts that are not from natural sciences (explanations) the tumultuous scene is palpable. This thesiscontributes to the field from two different perspectives. A theoretical one and a practical one. The formeris based on a profound literature review that resulted in two main contributions: 1) the proposition of anew definition for Explainable Artificial Intelligence and 2) the creation of a new taxonomy for the field.The latter is composed of two XAI frameworks that accommodate in some of the raging gaps found field,namely: 1) XAI framework for Echo State Networks and 2) XAI framework for the generation ofcounterfactual. The first accounts for the gap concerning Randomized neural networks since they havenever been considered within the field of XAI. Unfortunately, choosing the right parameters to initializethese reservoirs falls a bit on the side of luck and past experience of the scientist and less on that of soundreasoning. The current approach for assessing whether a reservoir is suited for a particular task is toobserve if it yields accurate results, either by handcrafting the values of the reservoir parameters or byautomating their configuration via an external optimizer. All in all, this poses tough questions to addresswhen developing an ESN for a certain application

Persistence is an important characteristic of many complex systems in nature, related to how long the system remains at a certain state before changing to a different one. The study of complex systems' persistence involves different definitions and uses different techniques, depending on whether short-term or long-term persistence is considered. In this paper we discuss the most important definitions, concepts, methods, literature and latest results on persistence in complex systems. Firstly, the most used definitions of persistence in short-term and long-term cases are presented. The most relevant methods to characterize persistence are then discussed in both cases. A complete literature review is also carried out. We also present and discuss some relevant results on persistence, and give empirical evidence of performance in different detailed case studies, for both short-term and long-term persistence. A perspective on the future of persistence concludes the work.

Magnetic resonance imaging (MRI) is an important non-invasive clinical tool that can produce high -resolution and reproducible images. However, a long scanning time is required for high-quality MR images, which leads to exhaustion and discomfort of patients, inducing more artefacts due to voluntary movements of the patients and involuntary physiological movements. To accelerate the scanning process, methods by k-space undersampling and deep learning based reconstruction have been popularised. This work introduced SwinMR, a novel Swin transformer based method for fast MRI reconstruction. The whole network consisted of an input module (IM), a feature extraction module (FEM) and an output module (OM). The IM and OM were 2D convolutional layers and the FEM was composed of a cascaded of residual Swin transformer blocks (RSTBs) and 2D convolutional layers. The RSTB consisted of a series of Swin transformer layers (STLs). The shifted windows multi-head self-attention (W-MSA/SW-MSA) of STL was performed in shifted windows rather than the multi-head self-attention (MSA) of the original trans -former in the whole image space. A novel multi-channel loss was proposed by using the sensitivity maps, which was proved to reserve more textures and details. We performed a series of comparative studies and ablation studies in the Calgary-Campinas public brain MR dataset and conducted a downstream seg-mentation experiment in the Multi-modal Brain Tumour Segmentation Challenge 2017 dataset. The results demonstrate our SwinMR achieved high-quality reconstruction compared with other benchmark methods, and it shows great robustness with different undersampling masks, under noise interruption and on different datasets. The code is publicly available at https://github.com/ayanglab/SwinMR.

Removing the bias and variance of multicentre data has always been a challenge in large scale digital healthcare studies, which requires the ability to integrate clinical features extracted from data acquired by different scanners and protocols to improve stability and robustness. Previous studies have described various computational approaches to fuse single modality multicentre datasets. However, these surveys rarely focused on evaluation metrics and lacked a checklist for computational data harmonisation studies. In this systematic review, we summarise the computational data harmonisation approaches for multi-modality data in the digital healthcare field, including harmonisation strategies and evaluation metrics based on different theories. In addition, a comprehensive checklist that summarises common practices for data harmonisation studies is proposed to guide researchers to report their research findings more effectively. Last but not least, flowcharts presenting possible ways for methodology and metric selection are proposed and the limitations of different methods have been surveyed for future research.

In clinical medicine, magnetic resonance imaging (MRI) is one of the most important tools for diagnosis, triage, prognosis, and treatment planning. However, MRI suffers from an inherent slow data acquisition process because data is collected sequentially in k-space. In recent years, most MRI reconstruction methods proposed in the literature focus on holistic image reconstruction rather than enhancing the edge information. This work steps aside this general trend by elaborating on the enhancement of edge information. Specifically, we introduce a novel parallel imaging coupled dual discriminator generative adversarial network (PIDD-GAN) for fast multi-channel MRI reconstruction by incorporating multi-view information. The dual discriminator design aims to improve the edge information in MRI reconstruction. One discriminator is used for holistic image reconstruction, whereas the other one is responsible for enhancing edge information. An improved U-Net with local and global residual learning is proposed for the generator. Frequency channel attention blocks (FCA Blocks) are embedded in the generator for incorporating attention mechanisms. Content loss is introduced to train the generator for better reconstruction quality. We performed comprehensive experiments on Calgary-Campinas public brain MR dataset and compared our method with state-of-the-art MRI reconstruction methods. Ablation studies of residual learning were conducted on the MICCAI13 dataset to validate the proposed modules. Results show that our PIDD-GAN provides high-quality reconstructed MR images, with well-preserved edge information. The time of single-image reconstruction is below 5ms, which meets the demand of faster processing.

Scene understanding plays a crucial role in autonomous driving by utilizing sensory data for contextual information extraction and decision making. Beyond modeling advances, the enabler for vehicles to become aware of their surroundings is the availability of visual sensory data, which expand the vehicular perception and realizes vehicular contextual awareness in real-world environments. Research directions for scene understanding pursued by related studies include person/vehicle detection and segmentation, their transition analysis, lane change, and turns detection, among many others Unfortunately, these tasks seem insufficient to completely develop fully-autonomous vehicles i.e. achieving level-5 autonomy, travelling just like human-controlled cars. This latter statement is among the conclusions drawn from this review paper: scene understanding for autonomous driving cars using vision sensors still requires significant improvements. With this motivation, this survey defines, analyzes, and reviews the current achievements of the scene understanding research area that mostly rely on computationally complex deep learning models. Furthermore, it covers the generic scene understanding pipeline, investigates the performance reported by the state-of-the-art, informs about the time complexity analysis of avant garde modeling choices, and highlights major triumphs and noted limitations encountered by current research efforts. The survey also includes a comprehensive discussion on the available datasets, and the challenges that, even if lately confronted by researchers, still remain open to date. Finally, our work outlines future research directions to welcome researchers and practitioners to this exciting domain.

Advances in information and signal processing technologies have a significant impact on autonomous driving (AD), improving driving safety while minimizing the efforts of human drivers with the help of advanced artificial intelligence (AI) techniques. Recently, deep learning (DL) approaches have solved several real-world problems of complex nature. However, their strengths in terms of control processes for AD have not been deeply investigated and highlighted yet. This survey highlights the power of DL architectures in terms of reliability and efficient real-time performance and overviews state-of-the-art strategies for safe AD, with their major achievements and limitations. Furthermore, it covers major embodiments of DL along the AD pipeline including measurement, analysis, and execution, with a focus on road, lane, vehicle, pedestrian, drowsiness detection, collision avoidance, and traffic sign detection through sensing and vision-based DL methods. In addition, we discuss on the performance of several reviewed methods by using different evaluation metrics, with critics on their pros and cons. Finally, this survey highlights the current issues of safe DL-based AD with a prospect of recommendations for future research, rounding up a reference material for newcomers and researchers willing to join this vibrant area of Intelligent Transportation Systems.

Unsupervised anomaly discovery in stream data is a research topic with many practical applications. However, in many cases, it is not easy to collect enough training data with labeled anomalies for supervised learning of an anomaly detector in order to deploy it later for identification of real anomalies in streaming data. It is thus important to design anomalies detectors that can correctly detect anomalies without access to labeled training data. Our idea is to adapt the Online evolving Spiking Neural Network (OeSNN) classifier to the anomaly detection task. As a result, we offer an Online evolving Spiking Neural Network for Unsupervised Anomaly Detection algorithm (OeSNN-UAD), which, unlike OeSNN, works in an unsupervised way and does not separate output neurons into disjoint decision classes. OeSNN-UAD uses our proposed new two-step anomaly detection method. Also, we derive new theoretical properties of neuronal model and input layer encoding of OeSNN, which enable more effective and efficient detection of anomalies in our OeSNN-UAD approach. The proposed OeSNN-UAD detector was experimentally compared with state-of-the-art unsupervised and semi-supervised detectors of anomalies in stream data from the Numenta Anomaly Benchmark and Yahoo Anomaly Datasets repositories. Our approach outperforms the other solutions provided in the literature in the case of data streams from the Numenta Anomaly Benchmark repository. Also, in the case of real data files of the Yahoo Anomaly Benchmark repository, OeSNN-UAD outperforms other selected algorithms, whereas in the case of Yahoo Anomaly Benchmark synthetic data files, it provides competitive results to the results recently reported in the literature.

Advances in Data Science permeate every field of Transportation Science and Engineering, resulting in developments in the transportation sector that are data-driven. Nowadays, Intelligent Transportation Systems (ITS) could be arguably approached as a “story” intensively producing and consuming large amounts of data. A diversity of sensing devices densely spread over the infrastructure, vehicles or the travelers’ personal devices act as sources of data flows that are eventually fed into software running on automatic devices, actuators or control systems producing, in turn, complex information flows among users, traffic managers, data analysts, traffic modeling scientists, etc. These information flows provide enormous opportunities to improve model development and decision-making. This work aims to describe how data, coming from diverse ITS sources, can be used to learn and adapt data-driven models for efficiently operating ITS assets, systems and processes; in other words, for data-based models to fully become actionable. Grounded in this described data modeling pipeline for ITS, we define the characteristics, engineering requisites and challenges intrinsic to its three compounding stages, namely, data fusion, adaptive learning and model evaluation. We deliberately generalize model learning to be adaptive, since, in the core of our paper is the firm conviction that most learners will have to adapt to the ever-changing phenomenon scenario underlying the majority of ITS applications. Finally, we provide a prospect of current research lines within Data Science that can bring notable advances to data-based ITS modeling, which will eventually bridge the gap towards the practicality and actionability of such models.

[EN]Bio-inspired optimization (including Evolutionary Computation and Swarm Intelligence) is a growing research topic with many competitive bio-inspired algorithms being proposed every year. In such an active area, preparing a successful proposal of a new bio-inspired algorithm is not an easy task. Given the maturity of this research field, proposing a new optimization technique with innovative elements is no longer enough. Apart from the novelty, results reported by the authors should be proven to achieve a significant advance over previous outcomes from the state of the art. Unfortunately, not all new proposals deal with this requirement properly. Some of them fail to select appropriate benchmarks or reference algorithms to compare with. In other cases, the validation process carried out is not defined in a principled way (or is even not done at all). Consequently, the significance of the results presented in such studies cannot be guaranteed. In this work we review several recommendations in the literature and propose methodological guidelines to prepare a successful proposal, taking all these issues into account. We expect these guidelines to be useful not only for authors, but also for reviewers and editors along their assessment of new contributions to the field.

135 p., A sequence is a collection of data instances arranged in an structured manner. When thisarrangement is held in the time domain, sequences are instead referred to as time series. As such,each observation in a time series represents an observation drawn from an underlying process,produced at a specific time instant. However, other type of data indexing structures, such as spaceorthreshold-based arrangements are possible. Data points that compose a time series are oftencorrelated to each other. To account for this correlation in data mining tasks, time series are usuallystudied as a whole data object rather than as a collection of independent observations. In thiscontext, techniques for time series analysis aim at analyzing this type of data structures by applyingspecific approaches developed to harness intrinsic properties of the time series for a wide range ofproblems such as, classification, clustering and other tasks alike.The development of monitoring and storage devices has made time series analysisproliferate in numerous application fields including medicine, economics, manufacturing andtelecommunications, among others. Over the years, the community has gathered efforts towards thedevelopment of new data-based techniques for time series analysis suited to address the problemsand needs of such application fields. In the related literature, such techniques can be divided in threemain groups: feature-, model- and distance- based methods. The first group (feature-based)transforms time series into a collection of features, which are then used by conventional learningalgorithms to provide solutions to the task under consideration. In contrast, methods belonging to thesecond group (model-based) assume that each time series is drawn from a generative model, whichis then harnessed to elicit information from data. Finally, distance-based techniques operate directlyon raw time series. To this end, these latter methods resort to specially defined measures of distanceor simi

El objetivo de este proyecto consiste en el desarrollo de un sistema basado en algoritmos de Reinforcement Learning, que sea capaz de otorgar autonomía a un vehículo aéreo no tripulado, el cual deberá aprender a realizar un desplazamiento autónomo desde un punto origen, hasta un punto destino, sin llegar a colisionar con los obstáculos que se pueda encontrar a lo largo de su trayectoria. Para ello, se harán uso de las herramientas software necesarias, que permitan la simulación de escenarios tanto para el entrenamiento, como para la evaluación del algoritmo., Laburpena Proiektu honen helburua Reinforcement Learning-eko algoritmoetan oinarritutako sistema garatzea izango da, non gida gabeko aire ibilgailu bati jatorrizko puntu batetik helmuga punto baterainoko ibilbidea burutzeko autonomía emango zaio. Horrenbestez, ibilgailuak zehaztutako ibilbidea egiten ikasi beharko du aurki ditzakeen oztopoak saihestuz. Horretarako, entrenamendu zein ebaluazio inguruneak simulatzeko beharrezkoak diren software tresnak erabiliko dira., The aim of this Project is the development of a system based on Reinforcement Learning algorithms, which is capable of granting autonomy to an unmanned aerial vehicle, which must learn to make an autonomous displacement from an origin point, to a destination point, without colliding with the obstacles that can be found along its path. For this, the necessary software tools will be used, which allow the simulation of scenarios both for training and for the evaluation of the algorithm.

222 p., Nowadays there is little doubt that wireless communications have been a pivotal player in the irruption and maturity of digital technologies in almost all sectors of activity. Over the years, the society has witnessed how wireless networking has become an essential element of its evolution and prosperity. However, the ever-growing requirements of applications and services in terms of rate, reliability and latency have steered the interest of regulatory bodies towards emergent radio access interfaces capable of efficiently coping with such requisites. In this context, millimeter-wave (mmWave) communications have been widely acknowledged as a technology enabler for ultra-reliable, low-latency applications in forthcoming standards, such as 5G. Unfortunately, the unprecedented data rates delivered by mmWave communications come along with new paradigms in regards to radio resource allocation, user scheduling, and other issues all across the protocol stack, mainly due to the directivity of antennas andsensitiveness to blockage of communications held in this spectrum band. Consequently, the provision of machine intelligence to systems and processes relying on mmWave radio interfaces is a must for efficiently handling the aforementioned challenges.This Thesis contributes to the above research niche by exploring the use of elements and tools from Computational Intelligence, Matching Theory and Stochastic Optimization for the management of radio and network resources in mmWave communications. To this end, two different application scenarios are targeted: 1) Vehicular communications, where the high degree of mobility and the recurrent inter-vehicular blockage give rise to complex channel conditions for channel allocation and beam alignment; and 2) mobile virtual reality (VR), where the motion-to-photon latency limit raises the hurdle for scheduling the delivery of multimedia content over mmWave. A diversity of intelligent methods for clustering, predictive modeling, matching a

[ES] Resumen Las aplicaciones móviles son una fuente de vulnerabilidad para los hackers. Cada vez son más los ataques realizados a través de ellas. Por ello, es muy importante identificar qué aplicaciones son empleadas para realizar ataques. Esta identificación hace que el usuario evite la instalación de dichas aplicaciones en su dispositivo. Para poder clasificar una aplicación en malware o benignware, se crearán varios sistemas clasificadores mediante diferentes técnicas de Machine Learning. Para la creación de los sistemas, se emplearán técnicas tradicionales de aprendizaje basadas en algoritmos clasificadores y Deep Learning. De todos los sistemas creados con las técnicas tradicionales se elegirán tres de ellos: el que tenga mayor exactitud, el que presente la precisión más elevada y, por último, aquel que más sensibilidad tenga. Finalmente, para cada una de las tres métricas, se decidirá si elegir el sistema entrenado mediante aprendizaje profundo o el entrenado con el aprendizaje tradicional seleccionado anteriormente. Así, se podrá hacer uso de tres herramientas con distintos enfoques capaces de detectar aplicaciones malignas., [EU] Laburpena Aplikazio mugikorrak ahultasun jatorri bat dira hackerrentzat. Haien zehar egiten diren erasoak gero eta gehiago dira. Horregatik, oso garrantzitsua da erasoak egiteko erabiltzen diren aplikazioak ezagutzea. Hori jakitearekin, erabiltzaileak aplikazio horiek bere mugikorr gailuan ez instalatzea lortzen da. Aplikazio mugikor bat malware edo benignware sailkatu ahal izateko, hiru sistema sailkatzaile sortuko dira Machine Learning-eko teknika ezberdinak erabiliz. Sistemak sortzeko algoritmo sailkatzaileetan oinarritutako ohiko teknikak eta Deep Learning erabiliko dira. Ohiko teknikekin sortutako sistema guztietatik, hiru aukeratuko dira: doitasun garaiena duen sistema, sistema zehatzena eta sentikortasun handiena daukana. Azkenik, hiru metrika bakoitzerako, erabakiko da zein sistema aukeratu, sakon entrenamenduarekin sortutako sistema edo lehen aukeratu den ohikoz entrenatutakoa. Honela, aplikazio kaltegarriak sailkatzeko gai diren hiru tresna erabili ahal izango dira hiru ikuspegi ezberdinekin., [EN] Abstract Mobile applications are a source of vulnerability for hackers. Number of attacks performed through them are increasing. That is why It is really important to identify the applications which are mainly used to perform cyber-attacks. This app identification made the user avoid installing those apps on your mobile device. In order to be able to sort an app on malware or benignware, a classifier system will be set up using Machine Learning different methods. Shallow learning Techniques based on classifier algorithms and Deep Learning methods are going to be used so as to create the systems. Three of all of the systems created using shallow learning technics will be chosen: the one which holds the highest accuracy, other which holds the best precision and finally, that which holds the highest recall. In the end, depending on each metric, it will be decided whether to choose between the system trained by Deep Learning and the previously chosen one trained by Shallow Learning. Thereby, three tools capable of detecting malicious apps with different approaches will be available to users.

El crecimiento del uso de Internet, en concreto de las redes sociales, ha supuesto una complicación en la gestión de las mismas. En el campo de la minería de grafos, el cual nos permite visualizar redes con usuarios e interacciones entre los mismos, se ha tratado de buscar una forma de simplificación para ese análisis de datos. En este contexto, se plantea la búsqueda de comunidades que comparten atributos mediante algoritmos de optimización que alivien el coste computacional. En este trabajo, se presenta un algoritmo capaz de particionar una red de alta dimensionalidad de forma óptima, dando lugar a un avance en el campo de la optimización a gran escala, subrama de la heurística surgida recientemente que carece de avances relacionados con la minería de grafos., Interneteko erabileraren hazkundea, bereziki sare sozialak, konplikazio bat izan da horren kudeaketan. Grafikoen meatzaritzaren esparruan, erabiltzaileei eta horien arteko elkarrekintzak sarea bistaratzeko aukera ematen duena, datu horiek aztertzeko sinplifikazioa bilatzen hasi da. Testuinguru horretan, konputazionalaren kostua arintzen duten optimizazio algoritmoen bidez atributuak partekatzen dituzten komunitateen bilaketa sortzen da. Lan honetan, tamaina handiko dimentsionaltasun sare bat partizionatzeko gai den algoritmoa aurkezten da, eskala handiko optimizazioaren alorrean aurrerapena sortuz, duela gutxi grafoen meatzaritzarekin lotutako aurrerapenik ez duen heuristikaren azpialdean., The increase in Internet use, in particular social networks, has been a complication in the management of the same. In the field of graph mining, which allows us to visualize a network with users and interactions between them, we have begun to look for a simplification of them for the analysis of the data. In this context, the search for communities that share attributes is raised using optimization algorithms that alleviate the computational cost. This paper presents an algorithm capable of optimally dividing a high-dimensional network, leading to an advance in the field of large-scale optimization, a sub-branch of the recently emerged heuristic that lacks advances related to graph mining.

Resumen (castellano) Las paradas cardiorrespiratorias extrahospitalarias (PCREH) se posicionan como una de las principales causas de defunción en los países desarrollados. Ante dicho evento, existen ciertos factores determinantes para la supervivencia del sujeto, incluyendo la reanimación cardio pulmonar, una pronta desfibrilación y la calidad del tratamiento ofrecido por el Servicio de Emergencias Médicas. El corazón del paciente puede presentar hasta cinco tipos de ritmos distintos. Puesto que cada estado clínico precisa un tratamiento diferente, es de vital importancia para el personal médico, la pronta y correcta identificación del ritmo/estado del paciente. Por consiguiente, existen numerosos estudios dedicados al entendimiento de dichas patologías, los cuales emplean grabaciones de la señal electrocardiograma (ECG) durante episodios PCREH. Dichas grabaciones deben ser anotadas manualmente por un grupo de expertos clínicos. Por lo tanto, resulta una tarea dispendiosa, lo cual ocasiona escasez de bases de datos debidamente caracterizadas y anotadas. Con el objetivo de facilitar el acceso a colecciones de datos correctamente anotadas, existen algoritmos de anotación semiautomáticos. Estos algoritmos permiten identificar con elevada certeza, las patologías presentes en distintos intervalos temporales de la señal ECG. De esta forma, los expertos clínicos se focalizan en repasar las decisiones del algoritmo, ahorrando tiempo y coste. Por todo ello, los algoritmos de anotación facilitan los estudios de enfermedades cardiacas, mejorando la calidad del tratamiento realizado y, de esta forma, la probabilidad de supervivencia del paciente. En este trabajo se presentan cuatro clasificadores de ritmos de pacientes en PCREH. Para su desarrollo, primero se prepara una colección de episodios PCREH con los que entrenar los algoritmos. El primer clasificador extrae información únicamente de la señal ECG. El segundo añade la información presente en la impedancia transtorácica de, Summary (English) Out-of-hospital cardiac arrest (OHCA) is one of the leading causes of death in developed countries. There are several key factors to survive an OHCA event, including cardiopulmonary resuscitation, early defibrillation and the overall quality of treatment given by the Emergency Medical System. The patient's heart can present up to five different types of rhythms. Since each clinical condition requires a different treatment, a fast and precise identification of the patient's rhythm/status is crucial for the medical staff. Therefore, there are numerous studies that focus on the understanding of these pathologies, using electrocardiogram signals (ECG) recorded during OHCA events. These recordings must be manually annotated by a group of clinical experts. Because the high costs associated to manual annotation, there is a lack of properly characterized and annotated databases. In order to facilitate access to correctly annotated data collections, there are semiautomatic annotation algorithms. These algorithms identify with high accuracy the pathologies present in different time intervals of the ECG signal. In this way, clinical experts would focus on reviewing the algorithm's decisions, saving time and money. All these considerations make annotation algorithms a key factor to develop studies on OHCA, improving the quality of the treatment performed and the probability of patient survival. In this work, four classifiers of OHCA rhythms are presented. For their development, first a collection of OHCA episodes is prepared, in order to train the algorithms. The first classifier extracts information only from the ECG signal. The second one, adds the information present in the patient's transthoracic impedance. Then, a third classifier is developed using Deep Learning techniques, since it has recently demonstrated its potential in this field. After that, a fourth classifier is made optimizing the previous model. Lastly, the results are analysed and the perform, Laburpena (Euskara) Hospitalez kanpoko bihotz geldiketa (HKBG) mundo garatuko heriotza kausa handienetariko bat dira. Geldiketa bat ematen denean zenbait gertakari gako dira pazientearen biziraupenerako, adibidez bihotz biriketako masajea, desfibrilazio goiztiarra edota emergentzia zerbitzuek emandako tratamendua. Pazientearen bihotzak bost erritmo desberdin aurkez ditzazke HKGB batean. Egoera kliniko bakoitzak tratamendu desberdina behar duenez, pazientearen erritmoa/egoera goiz eta zehatz detektatzea oso garrantzitsua da. Ondorioz, lan asko egin dira patologia horiek ulertzeko eta identifikatzeko, orokorrean pazientearen grabatutako elektrokardiograma (EKG) erabiliz. Grabaketa horietan aditu klinikoek erritmoa identifikatu eta anotatu behar dute. Azken hau kostu handiko lana da, eta ondorioz oso HKBG datubase gutxi dago erritmo anotazio egokiekin. Erritmo anotazioak dituzten HKGB datubaseak sortzeko badira erritmoa modu erdiautomatikoan anotatzeko algoritmoak. Algoritmo hauek modu nahiko zehatzean identifika dezaketa HKGB pazientearen erritmo/egoera, horretarako grabatutako EKG erabiliz. Horrela aditu klinikoek emandako diagnostikoa baino ez dute berrikusi behar, denbora eta kostuak aurreztuz. Horregatik anotaziorako algoritmoek HKGBaren inguruko ikerkuntza errazteu eta hobetuko lituzkete, emandako tratamendua hobetuz, eta pazienteen biziraupen aukerak handituz. Lan honetan lau algoritmo garatu dira HKGB erritmoak modu automatikoan sailkatzeko. Algoritmoak garatzeko lehendabizi HKGB kasuen datubase bat prestatu da, algoritmoak entrenatzeko. Lehen sailkatzailea EKG-tik soilik lortzen du informazioa. Bigarrenak bular inpedantziako informazioa ere erabiltzen du. Gero, ikasketa sakonean oinarritutako sailkatzailea garatu da, esparru honetan teknika hauek oso emaitza onak eman izan baitituzte. Azkenik laugarren sailkatzailea aurrekoaren bertsio hobetua da. Bukatzeko, emaitzak aztertu eta sailkatzaileen errendimenduak alderatu dira.

RESUMEN CASTELLANO: El objetivo de este proyecto es desarrollar un sistema de exploración de grandes extensiones mediante una inteligencia de enjambre distribuida, basado en algoritmos de búsqueda de difusión estocástica. El enjambre está compuesto por varios agentes, tantos terrestres como voladores. Los cuales tendrán que comunicarse y cooperar entre sí para realizar tareas de exploración. Estás tareas consisten en delimitar zonas esparcidas por un mapa del cual los agentes no disponen de ninguna información previa. Adicionalmente, estas zonas requieren diferentes tipos de agentes y solo los correspondientes agentes podrán delimitarla. Estas especificaciones tratan de simular robots explorando grandes extensiones de terreno, las cuales contienen diversos obstáculos o peligros para el ser humano. Por ese motivo, los tipos de agentes representan robots con diferentes herramientas: cámaras, micrófonos, termómetros, barómetros, espectroscopio, etc. A la vez que simulan las características de esas zonas, las cuales podrían ser: fugas de gas, obstáculos físicos, peligros radioactivos, etc., The aim of this project is to develop asystem to explore huge land extensions with a distributed swarmintelligence. To accomplish this task Stochastic Diffusion Search algorithms will be used. The swarm is created by various agents, which can be land agents or flying agents.This agents will need to communicate and cooperate for accomplish exploration tasks.This tasks consist in demarcate areas scattered across a map. Additionally, this areas required different types of agents and only those types of agents can demarcate them.Robots exploring huge land extensions have beentrying to simulatewith these specifications. Moreover, the map contains several obstacles or dangers for the human being. In addition, the types of agents representvariousrobots with different tools:cameras, microphones, barometers, spectrometers and etcetera. At the same time, they also simulatethe zones, for example: gas leaks, physique obstacles or radioactive dangers., Proiektu honen helburu nagusia eremuhandien esplorazio sistema bat garatzea da.Banatutako taldeinteligentzia baten bidezetadifusio estokatikobilaketa algoritmoetan oinarrituta. Taldealurreko eta hegaleko hainbat agente ditu. Hauekkomunikatueta elkarlanean aritu beharko dira esploraziolanak egiteko. Zeregin hauek mapa batetan sakabanatuta dauden eremuak mugatzea da. Horrez gain, eremu horiek agente mota desberdinak eskatzen dituzte eta dagokion agenteak soilik mugatu ahal izango dituzte.Zehaztapen horiek eremu handiak esploratzen dituzten robotak simulatzen saiatzen dira, gizakiak oztopo edo arrisku ugari biltzen dituena. Hori dela eta, robotaktresna desberdinak dituzte: kamerak, termometroak, barometroak, espektroskopia, etab. Aldi berean, eremuenezaugarriak simulatzen dituzte, adibidez, gasen ihesak, oztopo fisikoak, arrisku erradioaktiboak,etab.

Sudden cardiac arrest is one of the leading causes of death in the industrialized world. Pulse detection is essential for the recognition of the arrest and the recognition of return of spontaneous circulation during therapy, and it is therefore crucial for the survival of the patient. This paper introduces the first method based exclusively on the ECG for the automatic detection of pulse during cardiopulmonary resuscitation. Random forest classifier is used to efficiently combine up to nine features from the time, frequency, slope, and regularity analysis of the ECG. Data from 191 cardiac arrest patients was used, and 1177 ECG segments were processed, 796 with pulse and 381 without pulse. A leave-one-patient out cross validation approach was used to train and test the algorithm. The statistical distributions of sensitivity (SE) and specificity (SP) for pulse detection were estimated using 500 patient-wise bootstrap partitions. The mean (std) SE/SP for nine-feature classifier was 88.4 (1.8) %/89.7 (1.4) %, respectively. The designed algorithm only requires 4-s-long ECG segments and could be integrated in any commercial automated external defibrillator. The method permits to detect the presence of pulse accurately, minimizing interruptions in cardiopulmonary resuscitation therapy, and could contribute to improve survival from cardiac arrest.

Este proyecto consiste en el desarrollo e implementación de un software que optimice el despliegue de nuevos nodos de comunicación en áreas catastróficas, para de esta forma proporcionar una mayor área de cobertura a los equipos de rescate. Para su desarrollo, se plantea un problema multi-objetivo (en el cual se busca el mínimo coste de despliegue y la máxima área cubierta) que dé como solución la posición de los nodos en un área extensa. Para la resolución de este problema se utilizaran algoritmos bioinspirado, los cuales proporcionan una solución equilibrada para los 2 objetivos planteados. En este problema se ha optado por los algoritmos NSGAII, SMPSO y MOEAD., Proiektu honek hondamendi eremuetan komunikazio-nodo berrien hedapena optimizatzen duen softwarea garatzea eta ezartzea du helburu, erreskate-taldeei estaldura-eremu handiagoa eskaintzeko. Garapenean zehar arazo anitzeko helburua proposatzen da (gutxieneko inplementazio-kostu eta estalitako gehineko azalera bilatuz), irtenbide gisa nodoen kokalekua ematen duenak eremu zabalean. Arazo honi aurre egiteko, proposatutako bi helburuak lortzeko bioinsipatuko algoritmoak erabiliko dira. Horretarako aukeratuko algoritmoak NSGAII, SMPSO eta MOEAD izan dira., This project consists in the development and implementation of a software that optimizes the deployment of new communication nodes in catastrophic areas, in order to provide a larger area of coverage for rescue teams. For its development, a multi-objective problem (in which the objectives are, provide the minimum cost of deployment and the maximum area covered) which gives as a result the position of the nodes in a large area. For the resolution of this problem bio-inspired algorithms will be used, these algorithms provide a balanced solution for the 2 objectives. In this problem we have opted for the following algorithms: NSGAII, SMPSO and MOEAD.

La mala gestión de los recursos disponibles es una de las principales causas de la rápida expansión de los incendios y también puede ocasionar víctimas. Esa es la razón por la cual es importante planificar los recursos disponibles, así como la geolocalización de cada unidad (la posición en la que cada unidad estará en cada momento). Por ello se creará una red neuronal que nos permita saber de antemano dónde estará cada unidad disponible. Para crear la red neuronal, se usarán diferentes tipos de regresores obteniendo una precisión de cada uno de ellos. Para concluir, se pretende decidir cuál de todos los métodos utilizados genera datos más precisos a partir de las predicciones obtenidas., Baliabide erabilgarrien kudeaketa okerra suteen hedapen azkarraren arrazoi nagusietako bat da eta hildakoak sor ditzake. Hori dela eta, eskuragarri dauden baliabideak eta unitate bakoitzaren geolokalizazioa (unitate bakoitza une bakoitzean egon dadin) planifikatzea garrantzitsua da. Horregatik, sare neuronal bat sortuko da eskuragarri dagoen unitate bakoitzerako posizioa aldez aurretik jakin dezagun. Sare neuronala sortzeko, erregresio mota ezberdinak erabiliko dira, haien zehaztasuna lortuz. Amaitzeko, erabilitako metodo guztien artean zeinek sortzen dituen iragarpenik zehatzenak zehaztuko da., The mismanagement of available resources is one of the main causes of the rapid expansion of fires and can also result in casualties. That is the reason why it is important to plan available resources, as well as the geolocation of each unit (the position where each unit will be at each moment). That is why a neural-network is going to be created, which will allow us to know in advanced where each available unit will be. To create the neural-network, different types of regressors are going to be used obtaining precision from each of them. To conclude, it is intended to decide which of all methods used for prediction generates more accurate data from the obtained precisions.

132 p., Road traffic management is a critical aspect for the design and planning of complex urban transport networks for which vehicle flow forecasting is an essential component. As a testimony of its paramount relevance in transport planning and logistics, thousands of scientific research works have covered the traffic forecasting topic during the last 50 years. In the beginning most approaches relied on autoregressive models and other analysis methods suited for time series data. During the last two decades, the development of new technology, platforms and techniques for massive data processing under the Big Data umbrella, the availability of data from multiple sources fostered by the Open Data philosophy and an ever-growing need of decision makers for accurate traffic predictions have shifted the spotlight to data-driven procedures. Even in this convenient context, with abundance of open data to experiment and advanced techniques to exploit them, most predictive models reported in literature aim for shortterm forecasts, and their performance degrades when the prediction horizon is increased. Long-termforecasting strategies are more scarce, and commonly based on the detection and assignment to patterns. These approaches can perform reasonably well unless an unexpected event provokes non predictable changes, or if the allocation to a pattern is inaccurate.The main core of the work in this Thesis has revolved around datadriven traffic forecasting, ultimately pursuing long-term forecasts. This has broadly entailed a deep analysis and understanding of the state of the art, and dealing with incompleteness of data, among other lesser issues. Besides, the second part of this dissertation presents an application outlook of the developed techniques, providing methods and unexpected insights of the local impact of traffic in pollution. The obtained results reveal that the impact of vehicular emissions on the pollution levels is overshadowed

153 p., Applications that generate data in the form of fast streams from non-stationary environments, that is,those where the underlying phenomena change over time, are becoming increasingly prevalent. In thiskind of environments the probability density function of the data-generating process may change overtime, producing a drift. This causes that predictive models trained over these stream data become obsoleteand do not adapt suitably to the new distribution. Specially in online learning scenarios, there is apressing need for new algorithms that adapt to this change as fast as possible, while maintaining goodperformance scores. Examples of these applications include making inferences or predictions based onfinancial data, energy demand and climate data analysis, web usage or sensor network monitoring, andmalware/spam detection, among many others.Online learning and concept drift are two of the most hot topics in the recent literature due to theirrelevance for the so-called Big Data paradigm, where nowadays we can find an increasing number ofapplications based on training data continuously available, named as data streams. Thus, learning in nonstationaryenvironments requires adaptive or evolving approaches that can monitor and track theunderlying changes, and adapt a model to accommodate those changes accordingly. In this effort, Iprovide in this thesis a comprehensive state-of-the-art approaches as well as I identify the most relevantopen challenges in the literature, while focusing on addressing three of them by providing innovativeperspectives and methods.This thesis provides with a complete overview of several related fields, and tackles several openchallenges that have been identified in the very recent state of the art. Concretely, it presents aninnovative way to generate artificial diversity in ensembles, a set of necessary adaptations andimprovements for spiking neural networks in order to be used in online learning scenarios, and finally, adrift detector based on

[ES]Este proyecto consiste en el diseño, implementación y despliegue de un software que pueda procesar muestras de datos generadas por dispositivos en tiempo real, con objeto de elaborar predicciones basadas en algoritmos de aprendizaje de máquina. Estas muestras responden a un problema de clasificación y están sujetas a un fenómeno denominado deriva de concepto, lo que implica que el patrón que se sigue para clasificarlas cambia a lo largo del tiempo. El software de aprendizaje debe ser capaz de tener en cuenta dicho cambio, detectándolo y adaptándose al mismo. Además, el proyecto se engloba en un escenario de Big Data, donde la cantidad de información a procesar es enorme. Por ello, el programa deberá implementarse en una plataforma especializada, que pueda ejecutarse en un clúster de computación con alta escalabilidad para procesar los datos en tiempo real, aprovechando los recursos disponibles en su totalidad., [EN]This project consists of the design, implementation and deployment of software that can process data samples generated by devices in real time to elaborate predictions based on machine learning algorithms. These samples belong to a classification problem and are subject to a phenomenon called concept drift, which implies that the pattern that was learnt to classify the samples may change over time. The learning software must be able to consider such change by detecting it and adapting to it. In addition, the project is set in a Big Data scenario, where the amount of information to be processed is enormous. Therefore, the program must be implemented in a specialized platform that can be deployed in a highly scalable computing cluster, to process the data in real time using the resources available in its entirety., [EU]Hurrengo proiektua denbora errealean gailu elektronikoen bidez sortutako datu erakusgarriak prozesatu ditzakeen software-aren diseinu, inplementazio eta hedatzean datza, makina ikaskuntzako algoritmotan oinarritutako iragarpenak lantzeko helburuarekin. Erakusgai hauek klasifikazio arazo bati erantzuten diote eta deribazio kontzeptuko izendatutako fenomeno bati lotuta daude. Honen ondorioz, klasifikazioa egiteko erabiltzen den patroia aldaketak izaten ditu denbora pasa ahala. Ikaskuntza software-a aipatutako aldaketak kontuan hartzeko gai izan behar da, hau antzematen eta beren arabera egoitzen. Bestalde, proiektua Big Data jokaleku baten barne hartzen da, prozesatu beharreko informazio kantitatea egundokoa denean. Horregatik, programa berezitutako plataforma batean inplementatu beharko da, non eskalabilizazio handiko konputazio cluster batean egikaritu daiteke datuak denbora errealean prozesatzeko, erabiltzeko moduan dauden baliabide osotasuna aprobetxatuz.

[ES]El informe expuesto a continuación presenta un primer estudio de que mutaciones o conjunto de mutaciones genéticas intervienen en el padecimiento de cáncer. El objetivo es diseñar un conjunto de módulos de SW que aporte las herramientas necesarias para poder determinar con un análisis genético las probabilidades reales del padecimiento de dicha enfermedad. El proyecto se realizará en colaboración de Tecnalia., [EU]Jarraian azaldutako informeak minbizia pairatzeko parte hartzen duten mutazio edo mutazio multzoen lehen azterketa aurkezten du. Helburua, SW moduluen multzoa diseinatzea da beharrezko tresnak bermatzeko. Horrela analisi genetikoaren bitartez aipatutako gaizotasuna jasateko probabilitate errealak zehazteko gai izan dadin. Proiektua Tecnaliaren laguntzarekin burutuko da., [EN]The following report presentes the first study about which genetic mutations or aggregate of mutations interfere in cancer condition. The goal is to design a set of SW modules that contributes with necessary tools in order to be able to determine the real probabilities of suffering from the disease by means of a genetic analysis

Telekomunikazio operadoreen hazkuntzarekin, bai operadore handiak, azpiegitura, antena eta irrati espektro propioak dituztenak, bai operadore birtualak zerbitzuak aurrekoei baliabideak alokatuz eskaintzen dituztenak, gaur egun, operadore asko daude azalera geografiko berean zerbitzuak ematen. Bakoitzak bere politika eta estrategia propioekin, bezeroei zein prezioei dagokionez. Gaur egungo merkatuan, operadore guztiak elkarrekin lehiatzen dute ahalik eta bezero kopuru handiena lortzeko asmotan eduki dezaketen errekurtso muga kontuan hartu gabe. Bezero bat operadorearen estaldura eremuan sartzen denean, automatikoki beharrezko baliabideak erreserbatzen dira berarentzat eta konexioa uneoro mantentzen da nahiz eta erabilera konstantea ez izan. Proiektu honetan, dispositibo edo gailuen konexio modu berri bat aurkezten da: operadoreak hasitako tethering-a. Tethering-a, errekurtsoen banaketa estrategia bat da, zeinak dispositibo bat tarte nodo bezala funtzionatzea ahalbidetzen duen. Modu honetan, operadoreak kudeatuta gailu bakarrarentzat errekurtso erreserba egin daiteke nahiz eta gehiago zerbitzatu. Gainera, content catching mekanismoak gehituta, tarteko dispositiboak cache bezala erabiltzea ahalbidetu dezake sareko edukiak bezeroengana hurbilduz. Proiektu honetan operadoreek euren artean aplikatzen dituzten erlazio-politika ezberdinen efektua aztertzen da, operadoreak izango dituen irabazietan eta tethering konexioen eraginez erabiltzaileak jasandako kalitate murrizketan. Politika hauek sortutako sarea eta emaitzak baldintzatzen dituzte, aztertutako bi parametroetan. Koalizio eta konpetentzia erlazioak zehaztuko dira hauei zuzenean erlazionatutako eta optimizazio multiobjetibo algoritmo baten bidez, Harmony Search, lortutako pareto-optimoak diren sare topologia ezberdinak proposatuko dira., Con el aumento en el número de operadores de telecomunicaciones, tanto grandes operadores que cuentan con infraestructuras propias y espectro de radio con licencia, como operadores virtuales que explotan las infraestructuras de los primeros para ofrecer servicios, actualmente varios ofrecen servicios simultáneamente en la misma área geográfica. Cada uno de ellos, con sus propias políticas y estrategias en cuanto a clientes y precios. En el mercado actual, todos los operadores compiten entre sí para conseguir el mayor número de clientes posible, sin tener en cuenta la limitación que puedan tener en los recursos disponibles, ya que cada usuario conectado a la red, supone reservar un ancho de banda específico y mantener esa conexión a pesar de que no esté en uso constante. En este proyecto, se plantea una nueva manera de conexión de dispositivos; tethering iniciado por el operador. El tethering es una estrategia de compartición de recursos, en la que se permite que un nodo se conecte a través de otro a la red, reservando así únicamente el ancho de banda necesario para uno de ellos. Además, añadiéndole el mecanismo content catching permite que el dispositivo intermedio pueda funcionar a modo de caché, acercando ciertos contenidos a los usuarios. En este proyecto, se estudia el efecto que las diferentes políticas relacionales de los operadores tienen tanto en los beneficios de los propios operadores como en la calidad de servicio recibida en los clientes. Estas políticas aplicadas, condicionarán los resultados obtenidos, tanto económicamente como en la degradación de la calidad de servicio debida al tethering. Se establecen relaciones de coalición y de competencia entre operadores y se proponen diferentes topologías Pareto-óptimas vinculadas a ellas, obtenidas mediante un algoritmo de optimización multiobjetivo: Harmony Seach., With the increase of the number of telecommunication operators, nowadays there are many operators offering services over the same geographical area, those operators can be big ones, who have physical infrastructure and spectrum license, or virtual operators who offer services exploiting the infrastructures of the big operators. Each one has its own politics and strategies regarding to clients and pricing. In the actual market, all the operators fight with each other for achieving the maximum number of clients as possible, without taking into account the physical resources limitations due to the maintenance of the connection of each user in spite of they aren’t using it constantly. In this Project a new connection way is laid out; operator started tethering connection. Tethering is a connection sharing strategy in which a node is allowed to reach the net thought other node’s connection saving this way the resources needed for one of the nodes. Besides, it allows the intermediate node work as a cache, bringing closer some contents to other users. In this work the application of bi-objective heuristics for the design of Pareto-optimal network topologies is studied depending of the different relational policies applied by the operators. This policies, will condition the obtained results, influencing the revenue of the incumbent operators in the scenario and the quality of service degradation experienced by the end users as a result of tethering Coalition and competition relationships are established between operators and different network topologies are proposed. These topologies will be obtained by the heuristic Harmony Search Algorithm.