Tracking recurrent concepts using context

The problem of recurring concepts in data stream classification is a special case of concept drift where concepts may reappear. Although several existing methods are able to learn in the presence of concept drift, few consider contextual information when tracking recurring concepts. Nevertheless, in many real-world scenarios context information is available and can be exploited to improve existing approaches in the detection or even anticipation of recurring concepts. In this work, we propose the extension of existing approaches to deal with the problem of recurring concepts by reusing previously learned decision models in situations where concepts reappear. The different underlying concepts are identified using an existing drift detection method, based on the error-rate of the learning process. A method to associate context information and learned decision models is proposed to improve the adaptation to recurring concepts. The method also addresses the challenge of retrieving the most appropriate concept for a particular context. Finally, to deal with situations of memory scarcity, an intelligent strategy to discard models is proposed. The experiments conducted so far, using synthetic and real datasets, show promising results and make it possible to analyze the trade-off between the accuracy gains and the learned models storage cost.

An approach to automatic learning assessment based on the computational theory of perceptions

E-learning systems output a huge quantity of data on a learning process. However, it takes a lot of specialist human resources to manually process these data and generate an assessment report. Additionally, for formative assessment, the report should state the attainment level of the learning goals defined by the instructor. This paper describes the use of the granular linguistic model of a phenomenon (GLMP) to model the assessment of the learning process and implement the automated generation of an assessment report. GLMP is based on fuzzy logic and the computational theory of perceptions. This technique is useful for implementing complex assessment criteria using inference systems based on linguistic rules. Apart from the grade, the model also generates a detailed natural language progress report on the achieved proficiency level, based exclusively on the objective data gathered from correct and incorrect responses. This is illustrated by applying the model to the assessment of Dijkstra’s algorithm learning using a visual simulation-based graph algorithm learning environment, called GRAPHs

Regularized model learning in EDAs for continuous and multi-objective optimization

Probabilistic modeling is the de�ning characteristic of estimation of distribution algorithms (EDAs) which determines their behavior and performance in optimization. Regularization is a well-known statistical technique used for obtaining an improved model by reducing the generalization error of estimation, especially in high-dimensional problems. `1-regularization is a type of this technique with the appealing variable selection property which results in sparse model estimations. In this thesis, we study the use of regularization techniques for model learning in EDAs. Several methods for regularized model estimation in continuous domains based on a Gaussian distribution assumption are presented, and analyzed from di�erent aspects when used for optimization in a high-dimensional setting, where the population size of EDA has a logarithmic scale with respect to the number of variables. The optimization results obtained for a number of continuous problems with an increasing number of variables show that the proposed EDA based on regularized model estimation performs a more robust optimization, and is able to achieve signi�cantly better results for larger dimensions than other Gaussian-based EDAs. We also propose a method for learning a marginally factorized Gaussian Markov random �eld model using regularization techniques and a clustering algorithm. The experimental results show notable optimization performance on continuous additively decomposable problems when using this model estimation method. Our study also covers multi-objective optimization and we propose joint probabilistic modeling of variables and objectives in EDAs based on Bayesian networks, speci�cally models inspired from multi-dimensional Bayesian network classi�ers. It is shown that with this approach to modeling, two new types of relationships are encoded in the estimated models in addition to the variable relationships captured in other EDAs: objectivevariable and objective-objective relationships. An extensive experimental study shows the e�ectiveness of this approach for multi- and many-objective optimization. With the proposed joint variable-objective modeling, in addition to the Pareto set approximation, the algorithm is also able to obtain an estimation of the multi-objective problem structure. Finally, the study of multi-objective optimization based on joint probabilistic modeling is extended to noisy domains, where the noise in objective values is represented by intervals. A new version of the Pareto dominance relation for ordering the solutions in these problems, namely �-degree Pareto dominance, is introduced and its properties are analyzed. We show that the ranking methods based on this dominance relation can result in competitive performance of EDAs with respect to the quality of the approximated Pareto sets. This dominance relation is then used together with a method for joint probabilistic modeling based on `1-regularization for multi-objective feature subset selection in classi�cation, where six di�erent measures of accuracy are considered as objectives with interval values. The individual assessment of the proposed joint probabilistic modeling and solution ranking methods on datasets with small-medium dimensionality, when using two di�erent Bayesian classi�ers, shows that comparable or better Pareto sets of feature subsets are approximated in comparison to standard methods.

Diffusion Gradient Temporal Difference for Cooperative Reinforcement Learning with Linear Function Approximation

We introduce a diffusion-based algorithm in which multiple agents cooperate to predict a common and global statevalue function by sharing local estimates and local gradient information among neighbors. Our algorithm is a fully distributed implementation of the gradient temporal difference with linear function approximation, to make it applicable to multiagent settings. Simulations illustrate the benefit of cooperation in learning, as made possible by the proposed algorithm.

Decentralized multi-tasks distribution in heterogeneous robot teams by means of ant colony optimization and learning automata

This paper focuses on the general problem of coordinating multiple robots. More specifically, it addresses the self-election of heterogeneous specialized tasks by autonomous robots. In this paper we focus on a specifically distributed or decentralized approach as we are particularly interested on decentralized solution where the robots themselves autonomously and in an individual manner, are responsible of selecting a particular task so that all the existing tasks are optimally distributed and executed. In this regard, we have established an experimental scenario to solve the corresponding multi-tasks distribution problem and we propose a solution using two different approaches by applying Ant Colony Optimization-based deterministic algorithms as well as Learning Automata-based probabilistic algorithms. We have evaluated the robustness of the algorithm, perturbing the number of pending loads to simulate the robot’s error in estimating the real number of pending tasks and also the dynamic generation of loads through time. The paper ends with a critical discussion of experimental results.

PUE attack detection in CWSN using collaboration and learning behavior

Cognitive Wireless Sensor Network (CWSN) is a new paradigm which integrates cognitive features in traditional Wireless Sensor Networks (WSNs) to mitigate important problems such as spectrum occupancy. Security in Cognitive Wireless Sensor Networks is an important problem because these kinds of networks manage critical applications and data. Moreover, the specific constraints of WSN make the problem even more critical. However, effective solutions have not been implemented yet. Among the specific attacks derived from new cognitive features, the one most studied is the Primary User Emulation (PUE) attack. This paper discusses a new approach, based on anomaly behavior detection and collaboration, to detect the PUE attack in CWSN scenarios. A nonparametric CUSUM algorithm, suitable for low resource networks like CWSN, has been used in this work. The algorithm has been tested using a cognitive simulator that brings important results in this area. For example, the result shows that the number of collaborative nodes is the most important parameter in order to improve the PUE attack detection rates. If the 20% of the nodes collaborates, the PUE detection reaches the 98% with less than 1% of false positives.

Learning Bayesian networks from data by the incremental compilation of new network polynomials

Probabilistic graphical models are a huge research field in artificial intelligence nowadays. The scope of this work is the study of directed graphical models for the representation of discrete distributions. Two of the main research topics related to this area focus on performing inference over graphical models and on learning graphical models from data. Traditionally, the inference process and the learning process have been treated separately, but given that the learned models structure marks the inference complexity, this kind of strategies will sometimes produce very inefficient models. With the purpose of learning thinner models, in this master thesis we propose a new model for the representation of network polynomials, which we call polynomial trees. Polynomial trees are a complementary representation for Bayesian networks that allows an efficient evaluation of the inference complexity and provides a framework for exact inference. We also propose a set of methods for the incremental compilation of polynomial trees and an algorithm for learning polynomial trees from data using a greedy score+search method that includes the inference complexity as a penalization in the scoring function.

Efficient Monte Carlo methods for multi-dimensional learning with classifier chains

Multi-dimensional classification (MDC) is the supervised learning problem where an instance is associated with multiple classes, rather than with a single class, as in traditional classification problems. Since these classes are often strongly correlated, modeling the dependencies between them allows MDC methods to improve their performance – at the expense of an increased computational cost. In this paper we focus on the classifier chains (CC) approach for modeling dependencies, one of the most popular and highest-performing methods for multi-label classification (MLC), a particular case of MDC which involves only binary classes (i.e., labels). The original CC algorithm makes a greedy approximation, and is fast but tends to propagate errors along the chain. Here we present novel Monte Carlo schemes, both for finding a good chain sequence and performing efficient inference. Our algorithms remain tractable for high-dimensional data sets and obtain the best predictive performance across several real data sets.

Online Learning-based Robust Visual Tracking for Autonomous Landing of Unmanned Aerial Vehicles

Autonomous landing is a challenging and important technology for both military and civilian applications of Unmanned Aerial Vehicles (UAVs). In this paper, we present a novel online adaptive visual tracking algorithm for UAVs to land on an arbitrary field (that can be used as the helipad) autonomously at real-time frame rates of more than twenty frames per second. The integration of low-dimensional subspace representation method, online incremental learning approach and hierarchical tracking strategy allows the autolanding task to overcome the problems generated by the challenging situations such as significant appearance change, variant surrounding illumination, partial helipad occlusion, rapid pose variation, onboard mechanical vibration (no video stabilization), low computational capacity and delayed information communication between UAV and Ground Control Station (GCS). The tracking performance of this presented algorithm is evaluated with aerial images from real autolanding flights using manually- labelled ground truth database. The evaluation results show that this new algorithm is highly robust to track the helipad and accurate enough for closing the vision-based control loop.

Online Learning-based Robust Visual Tracking for Autonomous Landing of Unmanned Aerial Vehicles

Autonomous landing is a challenging and important technology for both military and civilian applications of Unmanned Aerial Vehicles (UAVs). In this paper, we present a novel online adaptive visual tracking algorithm for UAVs to land on an arbitrary field (that can be used as the helipad) autonomously at real-time frame rates of more than twenty frames per second. The integration of low-dimensional subspace representation method, online incremental learning approach and hierarchical tracking strategy allows the autolanding task to overcome the problems generated by the challenging situations such as significant appearance change, variant surrounding illumination, partial helipad occlusion, rapid pose variation, onboard mechanical vibration (no video stabilization), low computational capacity and delayed information communication between UAV and Ground Control Station (GCS). The tracking performance of this presented algorithm is evaluated with aerial images from real autolanding flights using manually- labelled ground truth database. The evaluation results show that this new algorithm is highly robust to track the helipad and accurate enough for closing the vision-based control loop.

Improving flexible databases searches using machine learning

El objetivo principal de este proyecto ha sido introducir aprendizaje automático en la aplicación FleSe. FleSe es una aplicación web que permite realizar consultas borrosas sobre bases de datos nítidos. Para llevar a cabo esta función la aplicación utiliza unos criterios para definir los conceptos borrosos usados para llevar a cabo las consultas. FleSe además permite que el usuario cambie estas personalizaciones. Es aquí donde introduciremos el aprendizaje automático, de tal manera que los criterios por defecto cambien y aprendan en función de las personalizaciones que van realizando los usuarios. Los objetivos secundarios han sido familiarizarse con el desarrollo y diseño web, al igual que recordar y ampliar el conocimiento sobre lógica borrosa y el lenguaje de programación lógica Ciao-Prolog. A lo largo de la realización del proyecto y sobre todo después del estudio de los resultados se demuestra que la agrupación de los usuarios marca la diferencia con la última versión de la aplicación. Esto se basa en la siguiente idea, podemos usar un algoritmo de aprendizaje automático sobre las personalizaciones de los criterios de todos los usuarios, pero la gran diversidad de opiniones de los usuarios puede llevar al algoritmo a concluir criterios erróneos o no representativos. Para solucionar este problema agrupamos a los usuarios intentando que cada grupo tengan la misma opinión o mismo criterio sobre el concepto. Y después de haber realizado las agrupaciones usar el algoritmo de aprendizaje automático para precisar el criterio por defecto de cada grupo de usuarios. Como posibles mejoras para futuras versiones de la aplicación FleSe sería un mejor control y manejo del ejecutable plserver. Este archivo se encarga de permitir a la aplicación web usar el lenguaje de programación lógica Ciao-Prolog para llevar a cabo la lógica borrosa relacionada con las consultas. Uno de los problemas más importantes que ofrece plserver es que bloquea el hilo de ejecución al intentar cargar un archivo con errores y en caso de ocurrir repetidas veces bloquea todas las peticiones siguientes bloqueando la aplicación. Pensando en los usuarios y posibles clientes, sería también importante permitir que FleSe trabajase con bases de datos de SQL en vez de almacenar la base de datos en los archivos de Prolog. Otra posible mejora basarse en distintas características a la hora de agrupar los usuarios dependiendo de los conceptos borrosos que se van ha utilizar en las consultas. Con esto se conseguiría que para cada concepto borroso, se generasen distintos grupos de usuarios, los cuales tendrían opiniones distintas sobre el concepto en cuestión. Así se generarían criterios por defecto más precisos para cada usuario y cada concepto borroso.---ABSTRACT---The main objective of this project has been to introduce machine learning in the application FleSe. FleSe is a web application that makes fuzzy queries over databases with precise information, using defined criteria to define the fuzzy concepts used by the queries. The application allows the users to change and custom these criteria. On this point is where the machine learning would be introduced, so FleSe learn from every new user customization of the criteria in order to generate a new default value of it. The secondary objectives of this project were get familiar with web development and web design in order to understand the how the application works, as well as refresh and improve the knowledge about fuzzy logic and logic programing. During the realization of the project and after the study of the results, I realized that clustering the users in different groups makes the difference between this new version of the application and the previous. This conclusion follows the next idea, we can use an algorithm to introduce machine learning over the criteria that people have, but the problem is the diversity of opinions and judgements that exists, making impossible to generate a unique correct criteria for all the users. In order to solve this problem, before using the machine learning methods, we cluster the users in order to make groups that have the same opinion, and afterwards, use the machine learning methods to precise the default criteria of each users group. The future improvements that could be important for the next versions of FleSe will be to control better the behaviour of the plserver file, that cost many troubles at the beginning of this project and it also generate important errors in the previous version. The file plserver allows the web application to use Ciao-Prolog, a logic programming language that control and manage all the fuzzy logic. One of the main problems with plserver is that when the user uploads a file with errors, it will block the thread and when this happens multiple times it will start blocking all the requests. Oriented to the customer, would be important as well to allow FleSe to manage and work with SQL databases instead of store the data in the Prolog files. Another possible improvement would that the cluster algorithm would be based on different criteria depending on the fuzzy concepts that the selected Prolog file have. This will generate more meaningful clusters, and therefore, the default criteria offered to the users will be more precise.

Bivariate Krawtchouk polynomials: Inversion and connection problems with the NAVIMA algorithm

In this paper we present a recurrent procedure to solve an inversion problem for monic bivariate Krawtchouk polynomials written in vector column form, giving its solution explicitly. As a by-product, a general connection problem between two vector column of monic bivariate Krawtchouk families is also explicitly solved. Moreover, in the non monic case and also for Krawtchouk families, several expansion formulas are given, but for polynomials written in scalar form.

Formation of temporal-feature maps by axonal propagation of synaptic learning

Computational maps are of central importance to a neuronal representation of the outside world. In a map, neighboring neurons respond to similar sensory features. A well studied example is the computational map of interaural time differences (ITDs), which is essential to sound localization in a variety of species and allows resolution of ITDs of the order of 10 μs. Nevertheless, it is unclear how such an orderly representation of temporal features arises. We address this problem by modeling the ontogenetic development of an ITD map in the laminar nucleus of the barn owl. We show how the owl's ITD map can emerge from a combined action of homosynaptic spike-based Hebbian learning and its propagation along the presynaptic axon. In spike-based Hebbian learning, synaptic strengths are modified according to the timing of pre- and postsynaptic action potentials. In unspecific axonal learning, a synapse's modification gives rise to a factor that propagates along the presynaptic axon and affects the properties of synapses at neighboring neurons. Our results indicate that both Hebbian learning and its presynaptic propagation are necessary for map formation in the laminar nucleus, but the latter can be orders of magnitude weaker than the former. We argue that the algorithm is important for the formation of computational maps, when, in particular, time plays a key role.

Structure-activity relationships derived by machine learning: the use of atoms and their bond connectivities to predict mutagenicity by inductive logic programming.

We present a general approach to forming structure-activity relationships (SARs). This approach is based on representing chemical structure by atoms and their bond connectivities in combination with the inductive logic programming (ILP) algorithm PROGOL. Existing SAR methods describe chemical structure by using attributes which are general properties of an object. It is not possible to map chemical structure directly to attribute-based descriptions, as such descriptions have no internal organization. A more natural and general way to describe chemical structure is to use a relational description, where the internal construction of the description maps that of the object described. Our atom and bond connectivities representation is a relational description. ILP algorithms can form SARs with relational descriptions. We have tested the relational approach by investigating the SARs of 230 aromatic and heteroaromatic nitro compounds. These compounds had been split previously into two subsets, 188 compounds that were amenable to regression and 42 that were not. For the 188 compounds, a SAR was found that was as accurate as the best statistical or neural network-generated SARs. The PROGOL SAR has the advantages that it did not need the use of any indicator variables handcrafted by an expert, and the generated rules were easily comprehensible. For the 42 compounds, PROGOL formed a SAR that was significantly (P < 0.025) more accurate than linear regression, quadratic regression, and back-propagation. This SAR is based on an automatically generated structural alert for mutagenicity.

Constellations and the unsupervised learning of graphs

In this paper, we propose a novel method for the unsupervised clustering of graphs in the context of the constellation approach to object recognition. Such method is an EM central clustering algorithm which builds prototypical graphs on the basis of fast matching with graph transformations. Our experiments, both with random graphs and in realistic situations (visual localization), show that our prototypes improve the set median graphs and also the prototypes derived from our previous incremental method. We also discuss how the method scales with a growing number of images.