No description available.

Despite the number of approaches established for Multiple Criteria Optimisation Problems, few of them have been developed for the decision making process. This research work proposes a new methodology for the solution of optimisation problems that involve multiple criteria emphasising the Decision-Maker's (DM's) preferences model and the use of evolutionary computation techniques and fuzzy logic. The use of genetic algorithms (GAs) is of vital importance to the development of this research. The use of operations research (OR) techniques and decision analysis is also considered vital. The aim of this project is to provide a definition of hybrid approaches that combine the strengths of GA and decision analysis. For this reason four hybrid models are proposed: 1. The GA-SEMOPS. 2. The fuzzy multiobjective genetic optimiser. 3. The GA-PROTRADE. 4. The interactive procedure for multiple objective optimisation problems. The main characteristics of these approaches are that they handle the DM's preferences in an interactive way and their objective functions are formulated using goal levels and surrogate functions. In order to demonstrate that these models can be used in different optimisation problems they have been applied to different case studies covering examples from environmental systems to land and human resource allocation. Each model was studied in depth, comparing the results found with those available in literature. In the majority of the cases, it was found that they performed better than existing methods. The investigations carried out showed that the proposed hybrid models can be considered as a very powerful tool for the solution of a wide variety of optimisation problems in situations from business to science and engineering.

The dissertation presents the investigation and development of novel adaptive computational techniques that provide a high level of performance when searching complex high-dimensional design spaces characterised by heavy non-linear constraint requirements. The objective is to develop a set of adaptive search engines that will allow the successful negotiation of such spaces to provide the design engineer with feasible high performance solutions. Constraint optimisation currently presents a major problem to the engineering designer and many attempts to utilise adaptive search techniques whilst overcoming these problems are in evidence. The most widely used method (which is also the most general) is to incorporate the constraints in the objective function and then use methods for unconstrained search. The engineer must develop and adjust an appropriate penalty function. There is no general solution to this problem neither in classical numerical optimisation nor in evolutionary computation. Some recent theoretical evidence suggests that the problem can only be solved by incorporating a priori knowledge into the search engine. Therefore, it becomes obvious that there is a need to classify constrained optimisation problems according to the degree of available or utilised knowledge and to develop search techniques applicable at each stage. The contribution of this thesis is to provide such a view of constrained optimisation, starting from problems that handle the constraints on the representation level, going through problems that have explicitly defined constraints (i.e., an easily computed closed form like a solvable equation), and ending with heavily constrained problems with implicitly defined constraints (incorporated into a single simulation model). At each stage we develop applicable adaptive search techniques that optimally exploit the degree of available a priori knowledge thus providing excellent quality of results and high performance. The proposed techniques are tested using both well known test beds and real world engineering design problems provided by industry.

No description available.

No description available.

No description available.

This thesis studies the introduction of a priori structure into the design of learning systems based on artificial neural networks applied to sequence recognition, in particular to phoneme recognition in continuous speech. Because we are interested in sequence analysis, algorithms for training recurrent networks are studied and an original algorithm for constrained recurrent networks is proposed and test results are reported. We also discuss the integration of connectionist models with other analysis tools that have been shown to be useful for sequences, such as dynamic programming and hidden Markov models. We introduce an original algorithm to perform global optimization of a neural network/hidden Markov model hybrid, and show how to perform such a global optimization on all the parameters of the system. Finally, we consider some alternatives to sigmoid networks: Radial Basis Functions, and a method for searching for better learning rules using a priori knowledge and optimization algorithms.

Competitive intelligence (CI) can be loosely defined as the process by which an organization legally and systematically collects, organizes, analyzes, and disseminates the information about its competitive environment. Notwithstanding the growing interest in CI, there are few empirical investigations on the work activities of CI professionals. This research addresses three basic questions: Who are CI professionals, which tasks and activities are they engaged in and how do they go about them, and what factors constrain their performance and completion of these tasks and activities? Twenty-eight CI professionals across Canada participated in the study from 24 different organizations, representing 16 specific industries. These CI professionals include various intelligence managers and analysts, market researchers, strategic advisors, and information specialists, representing two main groups: business professionals and information professionals. Their major goals are to heighten awareness of the competitive environment in which their organizations compete and to enhance decision making by their various clients. To achieve these goals, they engage in 10 general classes of activities: news scanning and monitoring; project management; responding ad hoc requests; communicating with various stakeholders; preparing CI products/deliverables; perusing and evaluating various materials; writing and editing diverse documents; coaching and training other staff for CI; undertaking training themselves; and administrative, non-CI, and sundry other activities. Among them, most of time is allocated to preparing CI products or deliverables, communicating with various stakeholders, and email processing and news scanning. Most of the information needs of the participants are not personal but derive from their organizational needs and clientele. The information seeking behavior of the participants can be situated on four axes: cyclical and noncyclical, reactive and proactive, linear and / Malgré l’intérêt grandissant pour la veille concurrentielle, il existe peu d’études empiriques traitant du travail des professionnels oeuvrant dans ce domaine. La présente étude pose trois questions de base: qui sont les professionnels faisant de la veille concurrentielle; quelles sont leurs tâches et activités et comment sont elles réalisées; et finalement, quels facteurs limitent leur performance et capacité de compléter ces tâches et ces activités. À travers le Canada, 28 professionnels de la veille concurrentielle ont été recrutés pour l’étude représentant 24 organisations différentes dans 12 industries générales et 16 de type spécifique. Ces professionnels représentent des analystes, des gestionnaires de l’information, des chercheurs en études de marché, des conseillers stratégiques et des spécialistes de l’information. Leurs objectifs principaux consistent à accroître la conscientisation à l’environnement concurrentiel dans lequel leurs organisations rivalisent et à rehausser la qualité de la prise de décision chez leurs différents clients. Pour atteindre ces objectifs, ils et elles s’impliquent dans dix catégories distinctes d’activités: scruter les nouvelles et gérer une veille informationnelle; gérer des projets; répondre aux demandes d’information ponctuelles; communiquer avec les différentes parties; élaborer des produits de veille concurrentielle; lire et évaluer le matériel diversifié; rédiger et réviser des documents variés; former des employés à la veille concurrentielle; recevoir une formation continue; et s’impliquer dans d’autres activités administratives non-reliées au domaine de la veille. La plupart du temps est consacré à la préparation des produits de veille concurrentielle, aux communications avec les diverses parties, à la gestion de courriel et à la scrutation des nouvelles. La plupart des besoins informationnels des participants ne sont pas individuels mais$

The notion of mobile agent, a software component that can move autonomously between the different nodes of a network is gaining wide popularity in business and in academia. The term mobile agent was first introduced in 1994. Since then, lots of research has been carried out in various aspects of the newly introduced paradigm. It might even be surprising to know that a recent census reports the existence of more than 70 mobile agent systems. Therefore, there is a need to gather and analyze what has been done so far in this new area. / This survey reviews the field of mobile agents by summarizing the key concepts and giving an overview of the most important implementations. Design and implementation issues of mobile agents are analyzed in general. Some of the most important mobile agent systems are presented and discussed. Java's support for mobile agent development is thoroughly examined. In addition, the role of the Common Object Request Broker Architecture (CORBA) as a broker between mobile agents and their environment is also analyzed. Most importantly, a survey of the major security concerns is provided followed by an analysis of the currently available techniques to address these concerns. Last but not least, a detailed analysis of the Foundation for Intelligent Physical Agents (FIPA) standards for interoperability between heterogeneous agents and their hosts is included. This survey will help in understanding the potentials of mobile agents and why they have not caught on. Once progress is made in the areas of security, programming language support for specific mobile agent requirements, and standards for coordination between heterogeneous agents, it is expected that the mobile agent paradigm will dramatically revolutionize the way the Internet is being used now.

There are basically three areas which can be explored to improve an HMM based recognizer, namely, parameter extraction, training methods and vocabulary representation. / The goal of parameter extraction is not only to find a compact and robust parametric representation of the speech signal, but also to find one which allows the HMMs to obtain the best possible recognition performance. Historically, improvements at this level have usually been obtained on a trial and error basis, using as much knowledge as possible about both the speech production and speech perception mechanisms. That is, the acoustic parameter extraction module has always been viewed as a separate module from the HMMs. This thesis will explore the concept of performing parameter extraction with a connectionist model, whose parameters can be learned from training data. / Two HMM training techniques are used in this thesis, namely MLE and MMIE. Parameter initialization, of critical importance for both, will be investigated for discrete, semi-continuous and continuous HMMs. Training processes involving a combination of MLE and MMIE training are studied. Other issues such as codebook exponents and the use of pause and silence models will also be explored. / Even if the vocabulary contains only 11 words, its representation is a very important issue. The effects of vocabulary representation with phoneme based, word based (with no sharing) and inter-word models will be experimentally evaluated. It will be shown how a word error rate of 0.23% and a string error rate of 0.68% can be achieved on the TIDIGITS corpus--a performance rivalling the best results ever reported by any group of researchers.