Embedding Emotional Context in Recommender Systems

Emotions are crucial for user's decision making in recommendation processes. We first introduce ambient recommender systems, which arise from the analysis of new trends on the exploitation of the emotional context in the next generation of recommender systems. We then explain some results of these new trends in real-world applications through the smart prediction assistant (SPA) platform in an intelligent learning guide with more than three million users. While most approaches to recommending have focused on algorithm performance. SPA makes recommendations to users on the basis of emotional information acquired in an incremental way. This article provides a cross-disciplinary perspective to achieve this goal in such recommender systems through a SPA platform. The methodology applied in SPA is the result of a bunch of technology transfer projects for large real-world rccommender systems

Estudio de la aplicación de sistemas basados en el conocimiento a la operación de una planta de tratamiento de residuos sólidos urbanos por valorización energética

Una de las actuaciones posibles para la gestión de los residuos sólidos urbanos es la valorización energética, es decir la incineración con recuperación de energía. Sin embargo es muy importante controlar adecuadamente el proceso de incineración para evitar en lo posible la liberación de sustancias contaminantes a la atmósfera que puedan ocasionar problemas de contaminación industrial.Conseguir que tanto el proceso de incineración como el tratamiento de los gases se realice en condiciones óptimas presupone tener un buen conocimiento de las dependencias entre las variables de proceso. Se precisan métodos adecuados de medida de las variables más importantes y tratar los valores medidos con modelos adecuados para transformarlos en magnitudes de mando. Un modelo clásico para el control parece poco prometedor en este caso debido a la complejidad de los procesos, la falta de descripción cuantitativa y la necesidad de hacer los cálculos en tiempo real. Esto sólo se puede conseguir con la ayuda de las modernas técnicas de proceso de datos y métodos informáticos, tales como el empleo de técnicas de simulación, modelos matemáticos, sistemas basados en el conocimiento e interfases inteligentes. En [Ono, 1989] se describe un sistema de control basado en la lógica difusa aplicado al campo de la incineración de residuos urbanos. En el centro de investigación FZK de Karslruhe se están desarrollando aplicaciones que combinan la lógica difusa con las redes neuronales [Jaeschke, Keller, 1994] para el control de la planta piloto de incineración de residuos TAMARA. En esta tesis se plantea la aplicación de un método de adquisición de conocimiento para el control de sistemas complejos inspirado en el comportamiento humano. Cuando nos encontramos ante una situación desconocida al principio no sabemos como actuar, salvo por la extrapolación de experiencias anteriores que puedan ser útiles. Aplicando procedimientos de prueba y error, refuerzo de hipótesis, etc., vamos adquiriendo y refinando el conocimiento, y elaborando un modelo mental. Podemos diseñar un método análogo, que pueda ser implementado en un sistema informático, mediante el empleo de técnicas de Inteligencia Artificial.Así, en un proceso complejo muchas veces disponemos de un conjunto de datos del proceso que a priori no nos dan información suficientemente estructurada para que nos sea útil. Para la adquisición de conocimiento pasamos por una serie de etapas: - Hacemos una primera selección de cuales son las variables que nos interesa conocer. - Estado del sistema. En primer lugar podemos empezar por aplicar técnicas de clasificación (aprendizaje no supervisado) para agrupar los datos y obtener una representación del estado de la planta. Es posible establecer una clasificación, pero normalmente casi todos los datos están en una sola clase, que corresponde a la operación normal. Hecho esto y para refinar el conocimiento utilizamos métodos estadísticos clásicos para buscar correlaciones entre variables (análisis de componentes principales) y así poder simplificar y reducir la lista de variables. - Análisis de las señales. Para analizar y clasificar las señales (por ejemplo la temperatura del horno) es posible utilizar métodos capaces de describir mejor el comportamiento no lineal del sistema, como las redes neuronales. Otro paso más consiste en establecer relaciones causales entre las variables. Para ello nos sirven de ayuda los modelos analíticos - Como resultado final del proceso se pasa al diseño del sistema basado en el conocimiento. El objetivo principal es aplicar el método al caso concreto del control de una planta de tratamiento de residuos sólidos urbanos por valorización energética. En primer lugar, en el capítulo 2 Los residuos sólidos urbanos, se trata el problema global de la gestión de los residuos, dando una visión general de las diferentes alternativas existentes, y de la situación nacional e internacional en la actualidad. Se analiza con mayor detalle la problemática de la incineración de los residuos, poniendo especial interés en aquellas características de los residuos que tienen mayor importancia de cara al proceso de combustión.En el capítulo 3, Descripción del proceso, se hace una descripción general del proceso de incineración y de los distintos elementos de una planta incineradora: desde la recepción y almacenamiento de los residuos, pasando por los distintos tipos de hornos y las exigencias de los códigos de buena práctica de combustión, el sistema de aire de combustión y el sistema de humos. Se presentan también los distintos sistemas de depuración de los gases de combustión, y finalmente el sistema de evacuación de cenizas y escorias.El capítulo 4, La planta de tratamiento de residuos sólidos urbanos de Girona, describe los principales sistemas de la planta incineradora de Girona: la alimentación de residuos, el tipo de horno, el sistema de recuperación de energía, y el sistema de depuración de los gases de combustión Se describe también el sistema de control, la operación, los datos de funcionamiento de la planta, la instrumentación y las variables que son de interés para el control del proceso de combustión.En el capítulo 5, Técnicas utilizadas, se proporciona una visión global de los sistemas basados en el conocimiento y de los sistemas expertos. Se explican las diferentes técnicas utilizadas: redes neuronales, sistemas de clasificación, modelos cualitativos, y sistemas expertos, ilustradas con algunos ejemplos de aplicación.Con respecto a los sistemas basados en el conocimiento se analizan en primer lugar las condiciones para su aplicabilidad, y las formas de representación del conocimiento. A continuación se describen las distintas formas de razonamiento: redes neuronales, sistemas expertos y lógica difusa, y se realiza una comparación entre ellas. Se presenta una aplicación de las redes neuronales al análisis de series temporales de temperatura.Se trata también la problemática del análisis de los datos de operación mediante técnicas estadísticas y el empleo de técnicas de clasificación. Otro apartado está dedicado a los distintos tipos de modelos, incluyendo una discusión de los modelos cualitativos.Se describe el sistema de diseño asistido por ordenador para el diseño de sistemas de supervisión CASSD que se utiliza en esta tesis, y las herramientas de análisis para obtener información cualitativa del comportamiento del proceso: Abstractores y ALCMEN. Se incluye un ejemplo de aplicación de estas técnicas para hallar las relaciones entre la temperatura y las acciones del operador. Finalmente se analizan las principales características de los sistemas expertos en general, y del sistema experto CEES 2.0 que también forma parte del sistema CASSD que se ha utilizado.El capítulo 6, Resultados, muestra los resultados obtenidos mediante la aplicación de las diferentes técnicas, redes neuronales, clasificación, el desarrollo de la modelización del proceso de combustión, y la generación de reglas. Dentro del apartado de análisis de datos se emplea una red neuronal para la clasificación de una señal de temperatura. También se describe la utilización del método LINNEO+ para la clasificación de los estados de operación de la planta.En el apartado dedicado a la modelización se desarrolla un modelo de combustión que sirve de base para analizar el comportamiento del horno en régimen estacionario y dinámico. Se define un parámetro, la superficie de llama, relacionado con la extensión del fuego en la parrilla. Mediante un modelo linealizado se analiza la respuesta dinámica del proceso de incineración. Luego se pasa a la definición de relaciones cualitativas entre las variables que se utilizan en la elaboración de un modelo cualitativo. A continuación se desarrolla un nuevo modelo cualitativo, tomando como base el modelo dinámico analítico.Finalmente se aborda el desarrollo de la base de conocimiento del sistema experto, mediante la generación de reglas En el capítulo 7, Sistema de control de una planta incineradora, se analizan los objetivos de un sistema de control de una planta incineradora, su diseño e implementación. Se describen los objetivos básicos del sistema de control de la combustión, su configuración y la implementación en Matlab/Simulink utilizando las distintas herramientas que se han desarrollado en el capítulo anterior.Por último para mostrar como pueden aplicarse los distintos métodos desarrollados en esta tesis se construye un sistema experto para mantener constante la temperatura del horno actuando sobre la alimentación de residuos.Finalmente en el capítulo Conclusiones, se presentan las conclusiones y resultados de esta tesis.

Distributed intelligent management of active networks

This paper focuses on improving computer network management by the adoption of artificial intelligence techniques. A logical inference system has being devised to enable automated isolation, diagnosis, and even repair of network problems, thus enhancing the reliability, performance, and security of networks. We propose a distributed multi-agent architecture for network management, where a logical reasoner acts as an external managing entity capable of directing, coordinating, and stimulating actions in an active management architecture. The active networks technology represents the lower level layer which makes possible the deployment of code which implement teleo-reactive agents, distributed across the whole network. We adopt the Situation Calculus to define a network model and the Reactive Golog language to implement the logical reasoner. An active network management architecture is used by the reasoner to inject and execute operational tasks in the network. The integrated system collects the advantages coming from logical reasoning and network programmability, and provides a powerful system capable of performing high-level management tasks in order to deal with network fault.

Unsupervised genetic algorithm deployed for intrusion detection

This paper represents the first step in an on-going work for designing an unsupervised method based on genetic algorithm for intrusion detection. Its main role in a broader system is to notify of an unusual traffic and in that way provide the possibility of detecting unknown attacks. Most of the machine-learning techniques deployed for intrusion detection are supervised as these techniques are generally more accurate, but this implies the need of labeling the data for training and testing which is time-consuming and error-prone. Hence, our goal is to devise an anomaly detector which would be unsupervised, but at the same time robust and accurate. Genetic algorithms are robust and able to avoid getting stuck in local optima, unlike the rest of clustering techniques. The model is verified on KDD99 benchmark dataset, generating a solution competitive with the solutions of the state-of-the-art which demonstrates high possibilities of the proposed method.

Wireless sensor networks

This paper provides an introduction to Wireless Sensor Networks (WSN), their applications in the field of control engineering and elsewhere and gives pointers to future research needs. WSN are collections of stand-alone devices which, typically, have one or more sensors (e.g. temperature, light level), some limited processing capability and a wireless interface allowing communication with a base station. As they are usually battery powered, the biggest challenge is to achieve the necessary monitoring whilst using the least amount of power.

Data based constructive identification - overcoming the curse of dimensionality

The modelling of a nonlinear stochastic dynamical processes from data involves solving the problems of data gathering, preprocessing, model architecture selection, learning or adaptation, parametric evaluation and model validation. For a given model architecture such as associative memory networks, a common problem in non-linear modelling is the problem of "the curse of dimensionality". A series of complementary data based constructive identification schemes, mainly based on but not limited to an operating point dependent fuzzy models, are introduced in this paper with the aim to overcome the curse of dimensionality. These include (i) a mixture of experts algorithm based on a forward constrained regression algorithm; (ii) an inherent parsimonious delaunay input space partition based piecewise local lineal modelling concept; (iii) a neurofuzzy model constructive approach based on forward orthogonal least squares and optimal experimental design and finally (iv) the neurofuzzy model construction algorithm based on basis functions that are Bézier Bernstein polynomial functions and the additive decomposition. Illustrative examples demonstrate their applicability, showing that the final major hurdle in data based modelling has almost been removed.

Robot free will

We introduce the perspex machine which unifies projective geometry and Turing computation and results in a supra-Turing machine. We show two ways in which the perspex machine unifies symbolic and non-symbolic AI. Firstly, we describe concrete geometrical models that map perspexes onto neural networks, some of which perform only symbolic operations. Secondly, we describe an abstract continuum of perspex logics that includes both symbolic logics and a new class of continuous logics. We argue that an axiom in symbolic logic can be the conclusion of a perspex theorem. That is, the atoms of symbolic logic can be the conclusions of sub-atomic theorems. We argue that perspex space can be mapped onto the spacetime of the universe we inhabit. This allows us to discuss how a robot might be conscious, feel, and have free will in a deterministic, or semi-deterministic, universe. We ground the reality of our universe in existence. On a theistic point, we argue that preordination and free will are compatible. On a theological point, we argue that it is not heretical for us to give robots free will. Finally, we give a pragmatic warning as to the double-edged risks of creating robots that do, or alternatively do not, have free will.

Learning in neural networks and stochastic approximation methods with averaging

The problem of adjusting the weights (learning) in multilayer feedforward neural networks (NN) is known to be of a high importance when utilizing NN techniques in various practical applications. The learning procedure is to be performed as fast as possible and in a simple computational fashion, the two requirements which are usually not satisfied practically by the methods developed so far. Moreover, the presence of random inaccuracies are usually not taken into account. In view of these three issues, an alternative stochastic approximation approach discussed in the paper, seems to be very promising.

Towards optimizing the selection of neurons in affordable neural networks

This paper considers variations of a neuron pool selection method known as Affordable Neural Network (AfNN). A saliency measure, based on the second derivative of the objective function is proposed to assess the ability of a trained AfNN to provide neuronal redundancy. The discrepancies between the various affordability variants are explained by correlating unique sub group selections with relevant saliency variations. Overall this study shows that the method in which neurons are selected from a pool is more relevant to how salient individual neurons are, than how often a particular neuron is used during training. The findings herein are relevant to not only providing an analogy to brain function but, also, in optimizing the way a neural network using the affordability method is trained.

Multiprobabilistic prediction in early medical diagnoses

This paper describes the methodology of providing multiprobability predictions for proteomic mass spectrometry data. The methodology is based on a newly developed machine learning framework called Venn machines. Is allows to output a valid probability interval. The methodology is designed for mass spectrometry data. For demonstrative purposes, we applied this methodology to MALDI-TOF data sets in order to predict the diagnosis of heart disease and early diagnoses of ovarian cancer and breast cancer. The experiments showed that probability intervals are narrow, that is, the output of the multiprobability predictor is similar to a single probability distribution. In addition, probability intervals produced for heart disease and ovarian cancer data were more accurate than the output of corresponding probability predictor. When Venn machines were forced to make point predictions, the accuracy of such predictions is for the most data better than the accuracy of the underlying algorithm that outputs single probability distribution of a label. Application of this methodology to MALDI-TOF data sets empirically demonstrates the validity. The accuracy of the proposed method on ovarian cancer data rises from 66.7 % 11 months in advance of the moment of diagnosis to up to 90.2 % at the moment of diagnosis. The same approach has been applied to heart disease data without time dependency, although the achieved accuracy was not as high (up to 69.9 %). The methodology allowed us to confirm mass spectrometry peaks previously identified as carrying statistically significant information for discrimination between controls and cases.

A Survey on Graphical Methods for Classification Predictive Performance Evaluation

Predictive performance evaluation is a fundamental issue in design, development, and deployment of classification systems. As predictive performance evaluation is a multidimensional problem, single scalar summaries such as error rate, although quite convenient due to its simplicity, can seldom evaluate all the aspects that a complete and reliable evaluation must consider. Due to this, various graphical performance evaluation methods are increasingly drawing the attention of machine learning, data mining, and pattern recognition communities. The main advantage of these types of methods resides in their ability to depict the trade-offs between evaluation aspects in a multidimensional space rather than reducing these aspects to an arbitrarily chosen (and often biased) single scalar measure. Furthermore, to appropriately select a suitable graphical method for a given task, it is crucial to identify its strengths and weaknesses. This paper surveys various graphical methods often used for predictive performance evaluation. By presenting these methods in the same framework, we hope this paper may shed some light on deciding which methods are more suitable to use in different situations.

Building binary-tree-based multiclass classifiers using separability measures

Various popular machine learning techniques, like support vector machines, are originally conceived for the solution of two-class (binary) classification problems. However, a large number of real problems present more than two classes. A common approach to generalize binary learning techniques to solve problems with more than two classes, also known as multiclass classification problems, consists of hierarchically decomposing the multiclass problem into multiple binary sub-problems, whose outputs are combined to define the predicted class. This strategy results in a tree of binary classifiers, where each internal node corresponds to a binary classifier distinguishing two groups of classes and the leaf nodes correspond to the problem classes. This paper investigates how measures of the separability between classes can be employed in the construction of binary-tree-based multiclass classifiers, adapting the decompositions performed to each particular multiclass problem. (C) 2010 Elsevier B.V. All rights reserved.

An investigation into feature construction to assist word sense disambiguation

Identifying the correct sense of a word in context is crucial for many tasks in natural language processing (machine translation is an example). State-of-the art methods for Word Sense Disambiguation (WSD) build models using hand-crafted features that usually capturing shallow linguistic information. Complex background knowledge, such as semantic relationships, are typically either not used, or used in specialised manner, due to the limitations of the feature-based modelling techniques used. On the other hand, empirical results from the use of Inductive Logic Programming (ILP) systems have repeatedly shown that they can use diverse sources of background knowledge when constructing models. In this paper, we investigate whether this ability of ILP systems could be used to improve the predictive accuracy of models for WSD. Specifically, we examine the use of a general-purpose ILP system as a method to construct a set of features using semantic, syntactic and lexical information. This feature-set is then used by a common modelling technique in the field (a support vector machine) to construct a classifier for predicting the sense of a word. In our investigation we examine one-shot and incremental approaches to feature-set construction applied to monolingual and bilingual WSD tasks. The monolingual tasks use 32 verbs and 85 verbs and nouns (in English) from the SENSEVAL-3 and SemEval-2007 benchmarks; while the bilingual WSD task consists of 7 highly ambiguous verbs in translating from English to Portuguese. The results are encouraging: the ILP-assisted models show substantial improvements over those that simply use shallow features. In addition, incremental feature-set construction appears to identify smaller and better sets of features. Taken together, the results suggest that the use of ILP with diverse sources of background knowledge provide a way for making substantial progress in the field of WSD.

On the influence of imputation in classification: practical issues

The substitution of missing values, also called imputation, is an important data preparation task for many domains. Ideally, the substitution of missing values should not insert biases into the dataset. This aspect has been usually assessed by some measures of the prediction capability of imputation methods. Such measures assume the simulation of missing entries for some attributes whose values are actually known. These artificially missing values are imputed and then compared with the original values. Although this evaluation is useful, it does not allow the influence of imputed values in the ultimate modelling task (e.g. in classification) to be inferred. We argue that imputation cannot be properly evaluated apart from the modelling task. Thus, alternative approaches are needed. This article elaborates on the influence of imputed values in classification. In particular, a practical procedure for estimating the inserted bias is described. As an additional contribution, we have used such a procedure to empirically illustrate the performance of three imputation methods (majority, naive Bayes and Bayesian networks) in three datasets. Three classifiers (decision tree, naive Bayes and nearest neighbours) have been used as modelling tools in our experiments. The achieved results illustrate a variety of situations that can take place in the data preparation practice.