Artificial Neural Networks (ANNs) have been developed for many applications but no detailed study has been made in the measure of their quality such as efficiency and complexity using appropriate metrics. Without an appropriate measurement, it is difficult to tell how an ANN performs on given applications. In addition, it is difficult to provide a measure of the algorithmic complexity of any given application. Further, it is difficult to make use of the results obtained in an application to predict the ANN's quality in a similar application. This research was undertaken to develop metrics, named Neural Metrics, that can be used in the measurement, construction and specification of backpropagation based supervised feedforward ANNs for applied science and engineering applications. A detailed analysis of backpropagation was carried out with a view to studying the mathematical definitions of the proposed metrics. Variants of backpropagation using various optimisation techniques were evaluated with similar computational and metric analysis. The research involved the evaluation of the proposed set of neural metrics using the computer implementation of training algorithms across a number of scientific and engineering benchmark problems including binary and real type training data. The result of the evaluation, for each type of problem, was a specification of values for all neural metrics and network parameters that can be used to successfully solve the same type of problem. With such a specification, neural users can reduce the uncertainty and hence time in choosing the appropriate network details for solving the same type of problem. It is also possible to use the specified neural metric values as reference points to further the experiments with a view to obtaining a better or sub-optimal solution for the problem. In addition, the generalised results obtained in this study provide users not only with a better understanding of the algorithmic complexity of the problem but also with a useful guideline on predicting the values of metrics that are normally determined empirically. It must be emphasised that this study only considers metrics for assessment of construction and off-line training of neural networks. The operational performance (e.g. on-line deployment of the trained networks) is outside the scope. Operational results (e.g. CPU time and run time errors) on training the networks off-line were obtained and discussed for each type of application problem.

M. Tech. Computer Science / This dissertation reports on the development of a new game playing technique based on Probabilistic Distance clustering (pd-clustering) method to evolve an Awale game player. Game playing is one classic and complex problems of artificial intelligence that has attracted the attention of researchers in computer science field of study.

This dissertation is concerned with the development of neural network-based methods to the control of robot manipulators and focusses on three different approaches for this purpose. In the first approach, an implementation of an intelligent adaptive control strategy in the execution of complex trajectory tracking tasks by using multilayer neural networks is demonstrated by exploiting the pattern classification capability of these nets. The network training is provided by a rule-based controller which is programmed to switch an appropriate adaptive control algorithm for each component type of motion constituting the overall trajectory tracking task. The second approach is based on the capability of trained neural networks for approximating input-output mappings. The use of dynamical networks with recurrent connections and efficient supervised training policies for the identification and adaptive control of a nonlinear process are discussed and a decentralized adaptive control strategy for a class of nonlinear dynamical systems with specific application to robotic manipulators is presented. An effective integration of the modelling of inverse dynamics property of neural nets with the robustness to unknown disturbances property of variable structure control systems is considered as the third approach. This methodology yields a viable procedure for selecting the control parameters adaptively and for designing a model-following adaptive control scheme for a class of nonlinear dynamical systems with application to robot manipulators.

published_or_final_version / Computer Science / Master / Master of Philosophy

Artificial Intelligence Lab, Department of MIS, University of Arizona / The results of a study that involved the creation of knowledge bases of concepts from large, operational textual databases are reported. Two East-bloc computing knowledge bases, both based on a semantic network structure, were created automatically using two statistical algorithms. With the help of four East-bloc computing experts, we evaluated the two knowledge bases in detail in a concept-association experiment based on recall and recognition tests. In the experiment, one of the knowledge bases that exhibited the asymmetric link property out-performed all four experts in recalling relevant concepts in East-bloc computing. The knowledge base, which contained about 20,O00 concepts (nodes) and 280,O00 weighted relationships (links), was incorporated as a thesaurus-like component into an intelligent retrieval system. The system allowed users to perform semantics-based information management and information retrieval via interactive, conceptual relevance feedback.

Artificial Intelligence Lab, Department of MIS, University of Arizona / This paper describes a study to explore the integration of Group Support Systems (GSS) and Artificial Intelligence (AI) technology to provide solicitation, analytical, visualization and sense-making support for attitudes from large distributed marketing focus groups. The paper describes two experiments and the concomitant evolutionary design and development of an attitude analysis process and the MindMine Comment Analysis Tool. The analysis process circumvents many of the problems associated with traditional data gathering via closed-ended questionnaires and potentially biased interviews by providing support for online free response evaluative comments. MindMine allows teams of raters to analyze comments from any source, including electronic meetings, discussion groups or surveys, whether they are Web-based or same-place. The analysis results are then displayed as visualizations that enable the team quickly to make sense of attitudes reflected in the comment set, which we believe provide richer information and a more detailed understanding of attitudes.

Artificial Intelligence Lab, Department of MIS, University of Arizona / With the exponential growth of the Internet, information can be shared among government agencies more easily than before. However, this also poses some design issues and challenges. This article reports on our experience in building an infrastructure for information sharing and collaboration in the law enforcement domain. Based on our user requirement studies with the Tucson Police Department, three main design challenges are identified and discussed in details. Based on our findings, we propose an infrastructure to address these issues. The proposed design consists of three modules, namely (1) Security and Confidentiality Management Module, (2) Information Access and Monitoring Module, and (3) Collaboration Module. A prototype system will be deployed and tested at the Tucson Police Department. We anticipate that our studies can potentially provide useful insight to other digital government research projects.

Artificial Intelligence Lab, Department of MIS, University of Arizona / We report results of an investigation where thirty subjects were observed performing subject-based search in an online catalog system. The observations have revealed a range of misconceptions users have when performing subject-based search. We have developed a taxonomy that characterizes these misconceptions and a knowledge representation which explains these misconceptions. Directions for improving search performance are also suggested.

Automated visual surveillance is one of the most actively researched areas in the past decade. Although current behaviour recognition systems provide us with a good understanding on the behaviour of individual moving objects present in an observed scene, they are not able to efficiently recognize the behaviour of groups formed by large numbers of moving objects. In this thesis, we present a HMM-based group behaviour recognition system which is capable of recognizing group behaviours effectively and efficiently. In our approach, we generate synthetic data for the training and validation of our behaviour recognition system. In addition, we use a single feature vector to represent the group dynamics, instead of using one feature vector for each pairwise interaction. Experimental results show accurate classification for both real-life data and simulated data from Lee's dataset. Therefore, we conclude that the proposed approach is a viable and accurate technique to perform group behaviour recognition in both simulated environment and real-life situations. Moreover, the high accuracy of the classification results obtained on real-life data, when only synthetic data was used for the training, suggests that it is possible to develop group behaviour models using synthetic data alone. / La surveillance visuelle automatisée est un domaine de recherche parmi les plus actifs au cours de la dernière décennie. Bien que les systèmes actuels de reconnaissance des comportements nous fournissent une bonne compréhension sur le comportement des objets en mouvement dans une scène observée, ils ne sont pas en mesure de reconnaître efficacement le comportement de groupes formés de plusieurs objets en mouvement. Dans cette mémoire, nous présentons un système de reconnaissance des comportements de groupes basé sur le modèle de Markov caché (MMC). Notre système est capable de reconnaître les comportements de groupe de façon efficace et efficiente. Dans notre approche, nous générons des données synthétiques pour former et valider notre système de reconnaissance des comportements. De plus, nous utilisons un vecteur caractéristique pour représenter la dynamique d'un groupe au lieu d'utiliser un vecteur pour chaque interaction entre deux objets en mouvement. Les résultats expérimentaux montrent une classification précise pour les données réelles et simulées utilisant la base de données de Lee. Par conséquent, nous concluons que l'approche proposée est une solution viable et une technique précise pour effectuer la reconnaissance des comportements de groupes dans un environnement simulé et dans des situations de la vie courante. Les résultats démontrent aussi qu'en utilisant uniquement des données synthétiques pour le former, le système classe avec une grande précision les comportements issues de situations réelles. Cela suggère qu'il est possible de développer des modèles de comportement de groupe en utilisant seulement les don

In this thesis, we applied both numerical methods and artificial intelligence techniques to NMR data processing and analysis. First, a comprehensive study of the Iterative Quadratic Maximum Likelihood (IQML) method applied to NMR spectral parameter estimation is reported. The IQML is compared to other conventional time domain data analysis methods. Extensive simulations demonstrate the superior performance of the IQML method. We also develop a new technique, which uses genetic algorithm with a priori knowledge, to improve the quantification of NMR spectral parameters. The new proposed method outperforms the other conventional methods, especially in the situations that there are signals close in frequencies and the signal-to-noise ratio of the FID is low. / The usefulness of Singular Value Decomposition (SVD) method in NMR data processing is further exploited. A new two dimensional spectral processing scheme based on SVD is proposed for suppressing strong diagonal peaks. The superior performance of this method is demonstrated on an experimental phase-sensitive COSY spectrum. / Finally, we studied the feasibility of using neural network predicted secondary structure information in the NMR data analysis. Protein chemical shift databases are compiled and are used with the neural network predicted protein secondary structure information to improve the accuracy of protein chemical shift prediction. The potential of this strategy for amino acid classification in NMR resonance assignment is explored.