On the impact of disproportional samples in credit scoring models: An application to a Brazilian bank data

Statistical methods have been widely employed to assess the capabilities of credit scoring classification models in order to reduce the risk of wrong decisions when granting credit facilities to clients. The predictive quality of a classification model can be evaluated based on measures such as sensitivity, specificity, predictive values, accuracy, correlation coefficients and information theoretical measures, such as relative entropy and mutual information. In this paper we analyze the performance of a naive logistic regression model (Hosmer & Lemeshow, 1989) and a logistic regression with state-dependent sample selection model (Cramer, 2004) applied to simulated data. Also, as a case study, the methodology is illustrated on a data set extracted from a Brazilian bank portfolio. Our simulation results so far revealed that there is no statistically significant difference in terms of predictive capacity between the naive logistic regression models and the logistic regression with state-dependent sample selection models. However, there is strong difference between the distributions of the estimated default probabilities from these two statistical modeling techniques, with the naive logistic regression models always underestimating such probabilities, particularly in the presence of balanced samples. (C) 2012 Elsevier Ltd. All rights reserved.

Structure learning of graphical models for task-oriented robot grasping

In the collective imaginaries a robot is a human like machine as any androids in science fiction. However the type of robots that you will encounter most frequently are machinery that do work that is too dangerous, boring or onerous. Most of the robots in the world are of this type. They can be found in auto, medical, manufacturing and space industries. Therefore a robot is a system that contains sensors, control systems, manipulators, power supplies and software all working together to perform a task. The development and use of such a system is an active area of research and one of the main problems is the development of interaction skills with the surrounding environment, which include the ability to grasp objects. To perform this task the robot needs to sense the environment and acquire the object informations, physical attributes that may influence a grasp. Humans can solve this grasping problem easily due to their past experiences, that is why many researchers are approaching it from a machine learning perspective finding grasp of an object using information of already known objects. But humans can select the best grasp amongst a vast repertoire not only considering the physical attributes of the object to grasp but even to obtain a certain effect. This is why in our case the study in the area of robot manipulation is focused on grasping and integrating symbolic tasks with data gained through sensors. The learning model is based on Bayesian Network to encode the statistical dependencies between the data collected by the sensors and the symbolic task. This data representation has several advantages. It allows to take into account the uncertainty of the real world, allowing to deal with sensor noise, encodes notion of causality and provides an unified network for learning. Since the network is actually implemented and based on the human expert knowledge, it is very interesting to implement an automated method to learn the structure as in the future more tasks and object features can be introduced and a complex network design based only on human expert knowledge can become unreliable. Since structure learning algorithms presents some weaknesses, the goal of this thesis is to analyze real data used in the network modeled by the human expert, implement a feasible structure learning approach and compare the results with the network designed by the expert in order to possibly enhance it.

Algoritmi esatti per il Job Shop Scheduling: approcci Mathematical Programming

In questa tesi ci occuperemo di fornire un modello MIP di base e di alcune sue varianti, realizzate allo scopo di comprenderne il comportamento ed eventualmente migliorarne l’efficienza. Le diverse varianti sono state costruite agendo in particolar modo sulla definizione di alcuni vincoli, oppure sui bound delle variabili, oppure ancora nell’obbligare il risolutore a focalizzarsi su determinate decisioni o specifiche variabili. Sono stati testati alcuni dei problemi tipici presenti in letteratura e i diversi risultati sono stati opportunamente valutati e confrontati. Tra i riferimenti per tale confronto sono stati considerati anche i risultati ottenibili tramite un modello Constraint Programming, che notoriamente produce risultati apprezzabili in ambito di schedulazione. Un ulteriore scopo della tesi è, infatti, comparare i due approcci Mathematical Programming e Constraint Programming, identificandone quindi i pregi e gli svantaggi e provandone la trasferibilità al modello raffrontato.

BDI agents for Real Time Strategy games

While the use of distributed intelligence has been incrementally spreading in the design of a great number of intelligent systems, the field of Artificial Intelligence in Real Time Strategy games has remained mostly a centralized environment. Despite turn-based games have attained AIs of world-class level, the fast paced nature of RTS games has proven to be a significant obstacle to the quality of its AIs. Chapter 1 introduces RTS games describing their characteristics, mechanics and elements. Chapter 2 introduces Multi-Agent Systems and the use of the Beliefs-Desires-Intentions abstraction, analysing the possibilities given by self-computing properties. In Chapter 3 the current state of AI development in RTS games is analyzed highlighting the struggles of the gaming industry to produce valuable. The focus on improving multiplayer experience has impacted gravely on the quality of the AIs thus leaving them with serious flaws that impair their ability to challenge and entertain players. Chapter 4 explores different aspects of AI development for RTS, evaluating the potential strengths and weaknesses of an agent-based approach and analysing which aspects can benefit the most against centralized AIs. Chapter 5 describes a generic agent-based framework for RTS games where every game entity becomes an agent, each of which having its own knowledge and set of goals. Different aspects of the game, like economy, exploration and warfare are also analysed, and some agent-based solutions are outlined. The possible exploitation of self-computing properties to efficiently organize the agents activity is then inspected. Chapter 6 presents the design and implementation of an AI for an existing Open Source game in beta development stage: 0 a.d., an historical RTS game on ancient warfare which features a modern graphical engine and evolved mechanics. The entities in the conceptual framework are implemented in a new agent-based platform seamlessly nested inside the existing game engine, called ABot, widely described in Chapters 7, 8 and 9. Chapter 10 and 11 include the design and realization of a new agent based language useful for defining behavioural modules for the agents in ABot, paving the way for a wider spectrum of contributors. Chapter 12 concludes the work analysing the outcome of tests meant to evaluate strategies, realism and pure performance, finally drawing conclusions and future works in Chapter 13.

Non-evolutive pattern recognition techniques: An application in medical image diagnostics

Lo studio dell’intelligenza artificiale si pone come obiettivo la risoluzione di una classe di problemi che richiedono processi cognitivi difficilmente codificabili in un algoritmo per essere risolti. Il riconoscimento visivo di forme e figure, l’interpretazione di suoni, i giochi a conoscenza incompleta, fanno capo alla capacità umana di interpretare input parziali come se fossero completi, e di agire di conseguenza. Nel primo capitolo della presente tesi sarà costruito un semplice formalismo matematico per descrivere l’atto di compiere scelte. Il processo di “apprendimento” verrà descritto in termini della massimizzazione di una funzione di prestazione su di uno spazio di parametri per un ansatz di una funzione da uno spazio vettoriale ad un insieme finito e discreto di scelte, tramite un set di addestramento che descrive degli esempi di scelte corrette da riprodurre. Saranno analizzate, alla luce di questo formalismo, alcune delle più diffuse tecniche di artificial intelligence, e saranno evidenziate alcune problematiche derivanti dall’uso di queste tecniche. Nel secondo capitolo lo stesso formalismo verrà applicato ad una ridefinizione meno intuitiva ma più funzionale di funzione di prestazione che permetterà, per un ansatz lineare, la formulazione esplicita di un set di equazioni nelle componenti del vettore nello spazio dei parametri che individua il massimo assoluto della funzione di prestazione. La soluzione di questo set di equazioni sarà trattata grazie al teorema delle contrazioni. Una naturale generalizzazione polinomiale verrà inoltre mostrata. Nel terzo capitolo verranno studiati più nel dettaglio alcuni esempi a cui quanto ricavato nel secondo capitolo può essere applicato. Verrà introdotto il concetto di grado intrinseco di un problema. Verranno inoltre discusse alcuni accorgimenti prestazionali, quali l’eliminazione degli zeri, la precomputazione analitica, il fingerprinting e il riordino delle componenti per lo sviluppo parziale di prodotti scalari ad alta dimensionalità. Verranno infine introdotti i problemi a scelta unica, ossia quella classe di problemi per cui è possibile disporre di un set di addestramento solo per una scelta. Nel quarto capitolo verrà discusso più in dettaglio un esempio di applicazione nel campo della diagnostica medica per immagini, in particolare verrà trattato il problema della computer aided detection per il rilevamento di microcalcificazioni nelle mammografie.

Automatic induction of lexical features

This thesis concerns artificially intelligent natural language processing systems that are capable of learning the properties of lexical items (properties like verbal valency or inflectional class membership) autonomously while they are fulfilling their tasks for which they have been deployed in the first place. Many of these tasks require a deep analysis of language input, which can be characterized as a mapping of utterances in a given input C to a set S of linguistically motivated structures with the help of linguistic information encoded in a grammar G and a lexicon L: G + L + C → S (1) The idea that underlies intelligent lexical acquisition systems is to modify this schematic formula in such a way that the system is able to exploit the information encoded in S to create a new, improved version of the lexicon: G + L + S → L' (2) Moreover, the thesis claims that a system can only be considered intelligent if it does not just make maximum usage of the learning opportunities in C, but if it is also able to revise falsely acquired lexical knowledge. So, one of the central elements in this work is the formulation of a couple of criteria for intelligent lexical acquisition systems subsumed under one paradigm: the Learn-Alpha design rule. The thesis describes the design and quality of a prototype for such a system, whose acquisition components have been developed from scratch and built on top of one of the state-of-the-art Head-driven Phrase Structure Grammar (HPSG) processing systems. The quality of this prototype is investigated in a series of experiments, in which the system is fed with extracts of a large English corpus. While the idea of using machine-readable language input to automatically acquire lexical knowledge is not new, we are not aware of a system that fulfills Learn-Alpha and is able to deal with large corpora. To instance four major challenges of constructing such a system, it should be mentioned that a) the high number of possible structural descriptions caused by highly underspeci ed lexical entries demands for a parser with a very effective ambiguity management system, b) the automatic construction of concise lexical entries out of a bulk of observed lexical facts requires a special technique of data alignment, c) the reliability of these entries depends on the system's decision on whether it has seen 'enough' input and d) general properties of language might render some lexical features indeterminable if the system tries to acquire them with a too high precision. The cornerstone of this dissertation is the motivation and development of a general theory of automatic lexical acquisition that is applicable to every language and independent of any particular theory of grammar or lexicon. This work is divided into five chapters. The introductory chapter first contrasts three different and mutually incompatible approaches to (artificial) lexical acquisition: cue-based queries, head-lexicalized probabilistic context free grammars and learning by unification. Then the postulation of the Learn-Alpha design rule is presented. The second chapter outlines the theory that underlies Learn-Alpha and exposes all the related notions and concepts required for a proper understanding of artificial lexical acquisition. Chapter 3 develops the prototyped acquisition method, called ANALYZE-LEARN-REDUCE, a framework which implements Learn-Alpha. The fourth chapter presents the design and results of a bootstrapping experiment conducted on this prototype: lexeme detection, learning of verbal valency, categorization into nominal count/mass classes, selection of prepositions and sentential complements, among others. The thesis concludes with a review of the conclusions and motivation for further improvements as well as proposals for future research on the automatic induction of lexical features.

A COMPARATIVE STUDY OF HEURISTIC OPTIMIZATION ALGORITHMS

Heuristic optimization algorithms are of great importance for reaching solutions to various real world problems. These algorithms have a wide range of applications such as cost reduction, artificial intelligence, and medicine. By the term cost, one could imply that that cost is associated with, for instance, the value of a function of several independent variables. Often, when dealing with engineering problems, we want to minimize the value of a function in order to achieve an optimum, or to maximize another parameter which increases with a decrease in the cost (the value of this function). The heuristic cost reduction algorithms work by finding the optimum values of the independent variables for which the value of the function (the “cost”) is the minimum. There is an abundance of heuristic cost reduction algorithms to choose from. We will start with a discussion of various optimization algorithms such as Memetic algorithms, force-directed placement, and evolution-based algorithms. Following this initial discussion, we will take up the working of three algorithms and implement the same in MATLAB. The focus of this report is to provide detailed information on the working of three different heuristic optimization algorithms, and conclude with a comparative study on the performance of these algorithms when implemented in MATLAB. In this report, the three algorithms we will take in to consideration will be the non-adaptive simulated annealing algorithm, the adaptive simulated annealing algorithm, and random restart hill climbing algorithm. The algorithms are heuristic in nature, that is, the solution these achieve may not be the best of all the solutions but provide a means to reach a quick solution that may be a reasonably good solution without taking an indefinite time to implement.

Symmetric inverse consistent nonlinear registration driven by mutual information.

A nonlinear viscoelastic image registration algorithm based on the demons paradigm and incorporating inverse consistent constraint (ICC) is implemented. An inverse consistent and symmetric cost function using mutual information (MI) as a similarity measure is employed. The cost function also includes regularization of transformation and inverse consistent error (ICE). The uncertainties in balancing various terms in the cost function are avoided by alternatively minimizing the similarity measure, the regularization of the transformation, and the ICE terms. The diffeomorphism of registration for preventing folding and/or tearing in the deformation is achieved by the composition scheme. The quality of image registration is first demonstrated by constructing brain atlas from 20 adult brains (age range 30-60). It is shown that with this registration technique: (1) the Jacobian determinant is positive for all voxels and (2) the average ICE is around 0.004 voxels with a maximum value below 0.1 voxels. Further, the deformation-based segmentation on Internet Brain Segmentation Repository, a publicly available dataset, has yielded high Dice similarity index (DSI) of 94.7% for the cerebellum and 74.7% for the hippocampus, attesting to the quality of our registration method.

Keeping citizens, companies, and organizations informed by Web KnowARR

For the main part, electronic government (or e-government for short) aims to put digital public services at disposal for citizens, companies, and organizations. To that end, in particular, e-government comprises the application of Information and Communications Technology (ICT) to support government operations and provide better governmental services (Fraga, 2002) as possible with traditional means. Accordingly, e-government services go further as traditional governmental services and aim to fundamentally alter the processes in which public services are generated and delivered, after this manner transforming the entire spectrum of relationships of public bodies with its citizens, businesses and other government agencies (Leitner, 2003). To implement this transformation, one of the most important points is to inform the citizen, business, and/or other government agencies faithfully and in an accessible way. This allows all the partaking participants of governmental affairs for a transition from passive information access to active participation (Palvia and Sharma, 2007). In addition, by a corresponding handling of the participants' data, a personalization towards these participants may even be accomplished. For instance, by creating significant user profiles as a kind of participants' tailored knowledge structures, a better-quality governmental service may be provided (i.e., expressed by individualized governmental services). To create such knowledge structures, thus known information (e.g., a social security number) can be enriched by vague information that may be accurate to a certain degree only. Hence, fuzzy knowledge structures can be generated, which help improve governmental-participants relationship. The Web KnowARR framework (Portmann and Thiessen, 2013; Portmann and Pedrycz, 2014; Portmann and Kaltenrieder, 2014), which I introduce in my presentation, allows just all these participants to be automatically informed about changes of Web content regarding a- respective governmental action. The name Web KnowARR thereby stands for a self-acting entity (i.e. instantiated form the conceptual framework) that knows or apprehends the Web. In this talk, the frameworks respective three main components from artificial intelligence research (i.e. knowledge aggregation, representation, and reasoning), as well as its specific use in electronic government will be briefly introduced and discussed.

Skolemization for Substructural Logics

The usual Skolemization procedure, which removes strong quantifiers by introducing new function symbols, is in general unsound for first-order substructural logics defined based on classes of complete residuated lattices. However, it is shown here (following similar ideas of Baaz and Iemhoff for first-order intermediate logics in [1]) that first-order substructural logics with a semantics satisfying certain witnessing conditions admit a “parallel” Skolemization procedure where a strong quantifier is removed by introducing a finite disjunction or conjunction (as appropriate) of formulas with multiple new function symbols. These logics typically lack equivalent prenex forms. Also, semantic consequence does not in general reduce to satisfiability. The Skolemization theorems presented here therefore take various forms, applying to the left or right of the consequence relation, and to all formulas or only prenex formulas.

When Does Neoadjuvant Chemotherapy Really Avoid Radiotherapy? Clinical Predictors of Adjuvant Radiotherapy in Cervical Cancer.

BACKGROUND The aim of this study was to identify clinical variables that may predict the need for adjuvant radiotherapy after neoadjuvant chemotherapy (NACT) and radical surgery in locally advanced cervical cancer patients. METHODS A retrospective series of cervical cancer patients with International Federation of Gynecology and Obstetrics (FIGO) stages IB2-IIB treated with NACT followed by radical surgery was analyzed. Clinical predictors of persistence of intermediate- and/or high-risk factors at final pathological analysis were investigated. Statistical analysis was performed using univariate and multivariate analysis and using a model based on artificial intelligence known as artificial neuronal network (ANN) analysis. RESULTS Overall, 101 patients were available for the analyses. Fifty-two (51 %) patients were considered at high risk secondary to parametrial, resection margin and/or lymph node involvement. When disease was confined to the cervix, four (4 %) patients were considered at intermediate risk. At univariate analysis, FIGO grade 3, stage IIB disease at diagnosis and the presence of enlarged nodes before NACT predicted the presence of intermediate- and/or high-risk factors at final pathological analysis. At multivariate analysis, only FIGO grade 3 and tumor diameter maintained statistical significance. The specificity of ANN models in evaluating predictive variables was slightly superior to conventional multivariable models. CONCLUSIONS FIGO grade, stage, tumor diameter, and histology are associated with persistence of pathological intermediate- and/or high-risk factors after NACT and radical surgery. This information is useful in counseling patients at the time of treatment planning with regard to the probability of being subjected to pelvic radiotherapy after completion of the initially planned treatment.

Classification of Satellite Images by means of Fuzzy Rules generated by a Genetic Algorithm

The data acquired by Remote Sensing systems allow obtaining thematic maps of the earth's surface, by means of the registered image classification. This implies the identification and categorization of all pixels into land cover classes. Traditionally, methods based on statistical parameters have been widely used, although they show some disadvantages. Nevertheless, some authors indicate that those methods based on artificial intelligence, may be a good alternative. Thus, fuzzy classifiers, which are based on Fuzzy Logic, include additional information in the classification process through based-rule systems. In this work, we propose the use of a genetic algorithm (GA) to select the optimal and minimum set of fuzzy rules to classify remotely sensed images. Input information of GA has been obtained through the training space determined by two uncorrelated spectral bands (2D scatter diagrams), which has been irregularly divided by five linguistic terms defined in each band. The proposed methodology has been applied to Landsat-TM images and it has showed that this set of rules provides a higher accuracy level in the classification process

Assessment of Hydrogen as suistinable clean energy

The progressive depletion of fossil fuels and their high contribution to the energy supply in this modern society forces that will be soon replaced by renewable fuels. But the dispersion and alternation of renewable energy production also undertake to reduce their costs to use as energy storage and hydrogen carrier. It is necessary to develop technologies for hydrogen production from all renewable energy storage technologies and the development of energy production from hydrogen fuel cells and cogeneration and tri generation systems. In order to propel this technological development discussed where the hydrogen plays a key role as energy storage and renewable energy, the National Centre of Hydrogen and Fuel Cell Technology Experimentation in Spain equipped with installations that enable scientific and technological design, develop, verify, certify, approve, test, measure and, more importantly, the facility ensures continuous operation for 24 hours a day, 365 days year. At the same time, the system is scalable so as to allow continuous adaptation of new technologies are developed and incorporated into the assembly to verify integration at the same time it checks the validity of their development. The transformation sector can be said to be the heart of the system, because without neglecting the other sectors, this should prove the validity of hydrogen as a carrier - energy storage are important efforts that have to do to demonstrate the suitability of fuel cells or internal combustion systems to realize the energy stored in hydrogen at prices competitive with conventional systems. The multiple roles to meet the fuel cells under different conditions of operation require to cover their operating conditions, many different sizes and applications. The fourth area focuses on integration is an essential complement within the installation. We must integrate not only the electricity produced, but also hydrogen is used and the heat generated in the process of using hydrogen energy. The energy management in its three forms: hydrogen chemical, electrical and thermal integration requires complicated and require a logic and artificial intelligence extremes to ensure maximum energy efficiency at the same time optimum utilization is achieved. Verification of the development and approval in the entire production system and, ultimately, as a demonstrator set to facilitate the simultaneous evolution of production technology, storage and distribution of hydrogen fuel cells has been assessed.