Дисертації з теми "Agent à usage général"

L’apprentissage par renforcement conventionnel implique l’entraînement d’un agent unimodal sur une tâche unique et bien définie, guidé par un signal de récompense optimisé pour le gradient. Ce cadre ne nous permet pas d’envisager un agent d’apprentissage adapté aux problèmes du monde réel impliquant des flux de diverses modalités, des tâches multiples, souvent mal définies, voire pas définies du tout. C’est pourquoi nous préconisons une transition vers un cadre plus général, visant à créer des algorithmes d’apprentissage par renforcement plus adaptables et intrinsèquement polyvalents. Pour progresser dans cette direction, nous identifions deux domaines d’intérêt principaux. Le premier est l’amélioration de l’exploration, qui permet à l’agent d’apprendre de l’environnement en dépendant le moins possible du signal de récompense. Nous présentons Latent Go-Explore (LGE), une généralisation de l’algorithme Go-Explore qui, malgré ses résultats impressionnants, était limité par une forte contrainte de connaissance du domaine. LGE atténue ces limitations et permet une application plus large dans un cadre plus général. LGE démontre son efficacité et sa polyvalence accrues en surpassant de manière significative les lignes de base dans tous les environnements testés. Le deuxième domaine d’intérêt est celui de la conception d’un agent polyvalent qui peut fonctionner dans une variété d’environnements, impliquant ainsi une structure multimodale et transcendant même le cadre séquentiel conventionnel de l’apprentissage par renforcement. Nous présentons Jack of All Trades (JAT), une architecture multimodale basée Transformers, spécialement conçue pour les tâches de décision séquentielle. En utilisant un seul ensemble de poids, JAT démontre sa robustesse et sa polyvalence, rivalisant avec son unique référence sur plusieurs benchmarks d’apprentissage par renforcement et montrant même des performances prometteuses sur des tâches de vision et textuelles. Nous pensons que ces deux contributions constituent une étape importante vers une approche plus générale de l’apprentissage par renforcement. En outre, nous présentons d’autres avancées méthodologiques et techniques qui sont étroitement liées à notre question de recherche initiale. La première est l’introduction d’un ensemble d’environnements robotiques simulés à récompense éparse, conçus pour fournir à la communauté les outils nécessaires à l’apprentissage dans des conditions de faible supervision. Trois ans après son introduction, cette contribution a été largement adoptée par la communauté et continue de faire l’objet d’une maintenance et d’un support actifs. D’autre part, nous présentons Open RL Benchmark, notre initiative pionnière visant à fournir un ensemble complet et entièrement enregistré d’expériences d’apprentissage par renforcement, allant au-delà des données typiques pour inclure toutes les métriques spécifiques à l’algorithme et au système. Ce benchmark vise à améliorer l’efficacité de la recherche en fournissant des données prêtes à l’emploi et en v vi facilitant la reproductibilité précise des expériences. Grâce à son approche communautaire, il est rapidement devenu une ressource importante, documentant plus de 25 000 exécutions. Ces avancées techniques et méthodologiques, associées aux contributions scientifiques décrites ci-dessus, visent à promouvoir une approche plus générale de l’apprentissage par renforcement et, nous l’espérons, représentent une étape significative vers le développement à terme d’un agent plus opérationnel
Conventional Reinforcement Learning (RL) involves training a unimodal agent on a single, well-defined task, guided by a gradient-optimized reward signal. This framework does not allow us to envisage a learning agent adapted to real-world problems involving diverse modality streams, multiple tasks, often poorly defined, sometimes not defined at all. Hence, we advocate for transitioning towards a more general framework, aiming to create RL algorithms that more inherently versatile.To advance in this direction, we identify two primary areas of focus. The first aspect involves improving exploration, enabling the agent to learn from the environment with reduced dependence on the reward signal. We present Latent Go-Explore (LGE), an extension of the Go-Explore algorithm. While Go-Explore achieved impressive results, it was constrained by domain-specific knowledge. LGE overcomes these limitations, offering wider applicability within a general framework. In various tested environments, LGE consistently outperforms the baselines, showcasing its enhanced effectiveness and versatility. The second focus is to design a general-purpose agent that can operate in a variety of environments, thus involving a multimodal structure and even transcending the conventional sequential framework of RL. We introduce Jack of All Trades (JAT), a multimodal Transformer-based architecture uniquely tailored to sequential decision tasks. Using a single set of weights, JAT demonstrates robustness and versatility, competing its unique baseline on several RL benchmarks and even showing promising performance on vision and textual tasks. We believe that these two contributions are a valuable step towards a more general approach to RL. In addition, we present other methodological and technical advances that are closely related to our core research question. The first is the introduction of a set of sparsely rewarded simulated robotic environments designed to provide the community with the necessary tools for learning under conditions of low supervision. Notably, three years after its introduction, this contribution has been widely adopted by the community and continues to receive active maintenance and support. On the other hand, we present Open RL Benchmark, our pioneering initiative to provide a comprehensive and fully tracked set of RL experiments, going beyond typical data to include all algorithm-specific and system metrics. This benchmark aims to improve research efficiency by providing out-of-the-box RL data and facilitating accurate reproducibility of experiments. With its community-driven approach, it has quickly become an important resource, documenting over 25,000 runs.These technical and methodological advances, along with the scientific contributions described above, are intended to promote a more general approach to Reinforcement Learning and, we hope, represent a meaningful step toward the eventual development of a more operative RL agent

L'apparition des limitations au niveau des interconnxions et, en particulier dans les supercalculateurs parallèles où les échanges de données deviennent prohibitifs, constitue une réelle butée aux performances globales des machines. Notre objectif est d'étuider une architecture de calculateur massivement parallèle sur la base d'un mariage idéal entre l'électronique et l'optique. L'architecture proposée se compose de processus électroniques (PEs) et utilise les atouts de l'optique pour les communications entre les PEs. Les applications visées conduisent au développement d'un réseau d'interconnexionx de type réseau multi-étage "shuffle-exchange" enrichi par des connxions de voisinage. Le réseau est validé par l'exécution s'applications telles que le produit matrice-vecteur et la transformée de Fourier à une et deux dimensions. Les possibilités d'intégration de la technologie électronique à moyen terme permet d'envisager un parallélisme de 128 x 128 PEs d'une complexité fonctionnelle équivalente à des microprocesseurs 8 bits (~5000 transistors). Une fréquence d'horloge acceptable pour les PEs pourrait être comprise entre 10 et 100 MHz. Le procédé de réalisation du réseau de communications inter-PEs proposé utilise des matrices de microlentilles et de microprismes. Son principe, validé expérimentalement, autorise un parallélisme 128 x 128.

The MyCiTi is currently generating large volumes of raw transactional information in the form of commuter smartcard transactions, which can be considered Big Data. Agent Based modelling (ABM) has been applied internationally as a means of deriving actionable intelligence from Big Data. It is proposed that ABM can be used to unlock the hidden potential within the aforementioned data. This paper demonstrates how to go about developing and calibrating a MATSim-based ABM to analyse AFC data. It is found that data formatting algorithms are critical in the preparation of data for modelling activities. These algorithms are highly complex, requiring significant time investment prior to development. Furthermore, the development of appropriate ABM calibration parameters requires careful consideration in terms of appropriate data collection, simulation testing, and justification. This study serves as strong evidence to suggest that ABM is an appropriate analysis technique for MyCiTi data systems. Validation exercises reveal that ABM is able to calculate on board bus usage and system behaviour with a strong degree of accuracy (R-squared 0.85). It is however recommended that additional research be conducted into more detailed calibration activities, such as fine-tuning agent behaviour during simulation. Ultimately this research study achieves its explorative objectives of model development and testing, and paves a way forward for future research into the practical applications of Big Data and ABM in the South African context.

Dans les années 1814-1870, la correspondance entre le procureur général de Lyon et le Garde des Sceaux montre le rôle méconnu du procureur : agent de surveillance et de propagande du régime. Il surveille tous les membres de la société judiciaire ; il intervient dans le déroulement des carrières ; il contrôle les citoyens et transmet ses informations au ministre pour éviter les insurrections dans son ressort : Ain, Loire et Rhône. Il surveille les agitateurs pour déceler les foyers de résistance. La presse, toujours surveillée, est parfois muselée. Il intervient pendant les campagnes électorales comme agent de propagande au profit des candidats officiels et participe activement à la répression des opposants, notamment dans les juridictions d'exception, mais aussi en poursuivant tous leurs actes politiques. Des moyens très précis ont été élaborés pour connaître les magistrats et la situation morale et politique de son ressort. Par ses discours, sa correspondance ou ses réquisitoires, il est l'agent particulier du Garde des Sceaux
From 1814 to 1870, the letters sent by the public prosecutor of Lyon to the chancellor show an unknown part of his mission : supervision and government's propaganda. He supervises the citizens and everyone who takes part in Justice; he interferes in the careers and gives all his informations to the government to avoid insurrections in his counties: Ain, Loire, Rhône. He keeps an eye on every opponent. Papers, always supervised, are sometimes muzzled. He takes part in march of elections and supports the official candidates. He represses actively the opponents, for example by taking part in outstanding jurisdictions. He knows everything on every magistrate and everything that happens in his counties, regarding political and moral situation thanks to more and more accurate methods. By his speeches, his letters and pleas, he is to be the particular Chancelor's agent

The research work presented in this thesis is directed at addressing the knowledge acquisition bottleneck frequently encountered in computer simulation. The central idea is to extract the required knowledge from video data and use this to drive a computer simulation instead of the more conventional approach of interviewing domain experts and somehow encapsulating this knowledge in a manner whereby it can be used in the context of computer simulation. More specifically the idea presented in this thesis is to extract object location information from video data and then to mine this information to identify Movement Patterns (MPs) and then to utalise these MPs in the context of computer simulation. To act as a focus for the work rodent behaviour simulation was considered. Partly because video data concerning rodent behaviour was relatively easy to obtain and partly because there is a genuine need to achieve a better understanding of rodent behaviour. This is especially the case in the context of crop damage. There are a variety of computer simulation frameworks. One that naturally lends itself to rodent simulation is Multi Agent Based Simulation (MABS) whereby the objects to be simulated (rodents) are encapsulated in terms of software agents. In more detail the work presented is directed at a number of research issues in the context of the above: (i) mechanisms to identify a moving object in video data and extracting associated location information, (ii) the mining of MPs from the extracted location information, (iii) the representation of MPs in such a way that they are compatible with computer simulation frameworks especially MABS frameworks and (iv) mechanisms where by MPs can be utilized and interacted with so as to drive a MABS. Overall two types of mechanisms are considered, Absolute and Relative. The operation of rodent MABSs, driven using the proposed MP concept, is fully illustrated in the context of different categories of scenarios. The evaluation of the proposed MP driven MABSs was conducted by comparing real world scenarios to parallel simulated scenarios. The results presented in the thesis demonstrated that the proposed mechanisms for extracting locations, and consequently mining MPs, from video data to drive a MABS provides a useful approach to effective computer simulation that will have wide ranging benefits.

In this thesis, a new reactive session reconstruction method '
Smart-SRA'
is introduced. Web usage mining is a type of web mining, which exploits data mining techniques to discover valuable information from navigations of Web users. As in classical data mining, data processing and pattern discovery are the main issues in web usage mining. The first phase of the web usage mining is the data processing phase including session reconstruction. Session reconstruction is the most important task of web usage mining since it directly affects the quality of the extracted frequent patterns at the final step, significantly. Session reconstruction methods can be classified into two categories, namely '
reactive'
and '
proactive'
with respect to the data source and the data processing time. If the user requests are processed after the server handles them, this technique is called as &lsquo
reactive&rsquo
, while in &lsquo
proactive&rsquo
strategies this processing occurs during the interactive browsing of the web site. Smart-SRA is a reactive session reconstruction techique, which uses web log data and the site topology. In order to compare Smart-SRA with previous reactive methods, a web agent simulator has been developed. Our agent simulator models behavior of web users and generates web user navigations as well as the log data kept by the web server. In this way, the actual user sessions will be known and the successes of different techniques can be compared. In this thesis, it is shown that the sessions generated by Smart-SRA are more accurate than the sessions constructed by previous heuristics.

Actuellement, le traitement des pathologies infectieuses chez les bovins et les animaux de compagnie est menacé par l’accroissement de l’antibiorésistance et des bactéries multirésistantes. Il est donc primordial pour les entreprises pharmaceutiques vétérinaires de développer de nouvelles gammes d’agents antibactériens spécifiques au domaine animale. Dans ce contexte, il a été montré que certains polyaminostérols naturels tels que la squalamine possèdent un fort potentiel antimicrobien. Ainsi, l’objectif de ce travail de thèse est de répondre à la problématique présentée par la société Virbac dans le traitement des mammites (bovins) et des otites/pyodermites (chien) par le développement d’une nouvelle classe de dérivés polyaminostéroïdiens synthétiques à large spectre antimicrobien. Pour cela, nous avons constitué une chimiothèque de composés originaux préparés à partir de différents acides biliaires selon des voies de synthèse inédites. Ces composés ont été évalués in vitro pour leur cytotoxicité et leurs activités antibactériennes contre diverses bactéries à Gram positif et à Gram négatif et nous avons pu établir la preuve de concept in vitro de leur potentiel thérapeutique en tant qu’agents antibactériens ou adjuvants d’antibiotiques. De plus, nous avons montré que l’un de ces nouveaux dérivés, la claramine A1, agit sur l’intégrité physique des membranes bactériennes et sur les performances d'efflux des pompes AcrAB-TolC. Ainsi, il apparaît finalement que ces dérivés polyaminostéroïdiens au mode d’action non conventionnel pourraient constituer une nouvelle classe d’agents antibactériens pour un usage en tant que substituts d’antibiotiques en médecine vétérinaire
Currently, the treatment of infectious pathologies in cattle and pets is threatened by the growing antimicrobial resistance and the development of multidrug-resistant bacteria. Thus it is necessary for the veterinary pharmaceutical firms to develop new lines of antibacterial agents. In this context, some natural polyaminosterols such as squalamine have gained interest due to their potent antimicrobial activities. Thus the aim of this PhD work is to provide an answer in the treatment of mastitis (cattle) and otitis/pyoderma (dog) by the development of a new class of synthetic polyaminosterols with a broad spectrum of antibacterial activity. In this purpose, a chemical library of original compounds has been prepared starting from various bile acids by using unprecedented synthesis procedures. These compounds were evaluated for their in vitro cytotoxicity on CHO cells as well as their antibacterial activities against Gram-positive and Gram-negative bacteria and we have establish the in vitro proof of concept of the therapeutic potential of this family of molecules as antibacterial agent or antibiotic adjuvant. Additional investigations were then conducted on one of these novel derivatives namely claramine A1 to deepen knowledge of its mechanism of action and showed that claramine A1 acts on the physical integrity of bacterial membranes and the efflux performance of AcrAB-TolC pumps. Based on the results of claramine A1, it finally appears that these new polyaminosterol derivatives possessing a non-classical mode of action pertain to a new class of antibacterial agents and could constitute a substitute for traditional antibiotics in veterinary medicine

Grâce au numérique, de nombreux dispositifs ont fait leur apparition dans les salles de classe et offrent ainsi la possibilité de multiplier les supports pédagogiques dans le processus d’enseignement/apprentissage. Le Tableau Blanc Interactif ou Tableau Numérique Interactif (TBI/TNI) en fait partie. Aussi, semble-t-il judicieux d’interroger les facteurs qui influencent l’effet d’un tel dispositif sur la performance scolaire, au regard des résultats contradictoires des études scientifiques sur la question.Ce travail de recherche doctorale est une recherche mixte, alliant à la fois des données quantitatives et qualitatives. A la suite de la revue de littérature, elle tente donc de répondre à la problématique suivante : Quels sont les facteurs qui influencent l’impact de l’usage du Tableau Blanc Interactif (TBI) sur la performance des élèves ?Avant la mise en œuvre de la recherche expérimentale, les hypothèses de départ présumaient que dans un processus d’enseignement/apprentissage médié par le TBI, la performance scolaire des élèves est influencée par le style d’apprentissage de l’élève d’une part, et d’autre part, qu’elle est fonction du type de motivation à l’égard du TBI.En vue de confronter ces hypothèses susmentionnées à la réalité du terrain, une procédure expérimentale de type pré-test/post-test, exécutée en deux phases, consistant à une situation d’enseignement/apprentissage sans le TBI et ensuite avec le TBI, a été menée, afin de mesurer la performance scolaire des élèves.Pour recueillir les données, les outils que sont le questionnaire, un test d’identification du style d’apprentissage selon le modèle VAK (Visuel, Auditif, Kinesthésique), les entretiens individuels semi-dirigés, une échelle de motivation et des observations in situ ont été utilisés.Suite à l’analyse des résultats obtenus à partir des alternatives non paramétriques du test de Khi2 et du test t de Student pour des échantillons appariés, des différences significatives sont observées au niveau de la performance des élèves de style d’apprentissage visuel, contrairement aux élèves ayant d’autres styles d’apprentissage majeur.L’impact de l’intervention du TBI dans le processus d’enseignement-apprentissage serait également fonction du genre et de la motivation de l’élève vis-à-vis du TBI. En effet, on constate que les élèves de sexe masculin et les élèves ayant une motivation autodéterminée à l’égard de l’outil technologique ont vu leurs performances augmenter significativement
Thanks to digital technology, many devices have appeared in classrooms and offer the possibility of multiplying teaching aids in the teaching / learning process. The interactive whiteboard (TBI / TNI) is one of them. Thus, it seems judicious to question the factors that influence the effect of such a device on school performance, given the contradictory results of scientific studies on the subject.This doctoral research work is a mixed research, combining both quantitative and qualitative data. As a result of the literature review, she tries to answer the following question: What are the factors that influence the impact of the use of the Interactive Whiteboard (IWB) on student performance?Prior to the implementation of experimental research, initial assumptions assumed that in an interactive whiteboard-mediated teaching / learning process, students' academic performance is influenced by the student's learning style of learning. on the one hand, and on the other, that it is a function of their degree of motivation with regard to the interactive whiteboard.In order to compare these above-mentioned hypotheses with the reality on the ground, a two-phase experimental procedure consisting of a teaching / learning situation without the interactive whiteboard and then with the interactive whiteboard was conducted in order to measure student academic performance.To collect the data, the tools that are the questionnaire, a test of identification of the learning style according to the model VAK (Visual, Auditory, Kinesthetic), the semi-directed individual interviews, a scale of motivation and observations in situ have been used.Following the analysis of the results obtained from the nonparametric alternatives of the Khi2 test and Student's t test for paired samples, significant differences are observed in the performance of visual learning style students, unlike students with other major learning styles.The impact of the intervention of the whiteboard in the teaching-learning process would also depend on the student's gender and motivation towards the whiteboard. Indeed, we find that male students and students with a self-determined motivation for the technological tool have seen their performance increase significantly

Ce travail vise à étudier le statut, c’est-à-dire la place et le traitement des termes scientifiques et techniques dans le dictionnaire général. On oppose généralement le dictionnaire général au dictionnaire spécialisé. Le premier s’intéresse aux mots (ou lexique de la langue générale) et le deuxième aux termes (ou lexique des langues spécialisées). Or, le dictionnaire général inclut, en plus des mots de la langue générale, un nombre important de termes. Ces termes figurant alors simultanément dans les deux types de dictionnaire font l’objet de cette étude. Celle-ci se propose ainsi d’examiner les deux aspects macrostructurel - choix et emplacement - et mircostructuel - étiquetage, définition, exemplification et, pour les dictionnaires bilingues, correspondants arabes - des termes dans ce type de dictionnaire. Les termes retenus dans cette recherche relèvent de domaines divers. Ils sont extraits d’un corpus trilingue anglais-français-arabe constitué de textes de vulgarisation publiés par l’ONU. Le statut de ces termes est analysé dans deux dictionnaires monolingues français, deux monolingues arabes et deux bilingues français-arabe
This research aims to study the status - place and treatment - of scientific and technical terms in general dictionaries. General dictionaries concentrate on words (or general language lexicon) whereas specialised dictionaries focus on terms (or specialised language lexicon). However, general dictionaries include, in addition to words, a large number of terms. These terms, which are simultaneously included in both dictionaries types, are the subject of this study. Thus, it proposes to examine the following aspects of the term: macrostructural - choice and location - and mircostructual - labelling, definition and exemplification as well as the term’s equivalent in the Arabic bilingual dictionaries. The terms which are used in this research are from various domains. They are extracted from a trilingual English-French-Arabic corpus which consists of texts published by the UN. The status of these terms is analysed according to two monolingual french dictionaries, two monolingual arabic dictionaries and two bilingual french-arabic dictionaries

Objective To describe off-label utilization of erythropoietic agents in ICU patients.Methods A retrospective, observational study design was used to describe off-label utilization of erythropoietin alfa and darbepoetin alfa by 91,357 patients in the ICU during January 2002- June 2004 at 433 U.S. hospitals. Results Approximately 62% of patients received erythropoietic agents off-label in the ICU compared to 49% in the non-ICU population. Off-label use in the ICU was more likely in teaching hospitals, in larger sized hospitals, in females, for certain physician specialties, and in different regions of the U.S., The longer the hospital stay, the less likely off-label prescribing occurred. ConclusionsOff-label use of erythropoietic agents was common in the ICU. Multiple factors affected off-label use in the ICU although no single one was dominant.

This master's thesis deals with proposal of a system for monitoring of user activities on workstations Windows XP over .NET 2.0 platform. The system should monitor running programs, processes and find out user activities in these programs like movement of the mouse or writing on the keyboard. A part of whole project is a suggestion of the system which contains specifications of requirements for this system, creation of the architecture, creation of important models, suggestion of the communication between separated part of system and final compile of results. The project should get information about user's activities and also information about usage of software installed on these workstations. The final system will be tested by use case study. Last but not least, project deals with ethic questions in the context of monitoring of users and discuss law questions about this project.

This thesis focuses on the genomic study of symbionts of two different groups of hymenopterans: bees and ants. Both groups of insects have major ecological impact, and investigating their microbiomes increases our understanding of their health, diversity and evolution. The study of the bee gut microbiome, including members of Lactobacillus and Bifidobacterium, revealed genomic processes related to the adaptation to the gut environment, such as the expansion of genes for carbohydrate metabolism and the acquisition of genes for interaction with the host. A broader genomic study of these genera demonstrated that some lineages evolve under strong and opposite substitution biases, leading to extreme GC content values. A comparison of codon usage patterns in these groups revealed ongoing shifts of optimal codons. In a separate study we analysed the genomes of several strains of Lactobacillus kunkeei, which inhabits the honey stomach of bees but is not found in their gut. We observed signatures of genome reduction and suggested candidate genes for host-interaction processes. We discovered a novel type of genome architecture where genes for metabolic functions are located in one half of the genome, whereas genes for information processes are located in the other half. This genome organization was also found in other Lactobacillus species, indicating that it was an ancestral feature that has since been retained. We suggest mechanisms and selective forces that may cause the observed organization, and describe processes leading to its loss in several lineages independently. We also studied the genome of a species of Rhizobiales bacteria found in ants. We discuss its metabolic capabilities and suggest scenarios for how it may affect the ants’ lifestyle. This genome contained a region with homology to the Bartonella gene transfer agent (GTA), which is a domesticated bacteriophage used to transfer bacterial DNA between cells. We propose that its unique behaviour as a specialist GTA, preferentially transferring host-interaction factors, originated from a generalist GTA that transferred random segments of chromosomal DNA. These bioinformatic analyses of previously uncharacterized bacterial lineages have increased our understanding of their physiology and evolution and provided answers to old and new questions in fundamental microbiology.

Made available in DSpace on 2016-08-17T14:52:57Z (GMT). No. of bitstreams: 1 Leandro Balby Marinho.pdf: 1280354 bytes, checksum: 9dd7494777864e76ff13f7880d7faa7b (MD5) Previous issue date: 2005-03-11
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
With the continuous migration of a great diversity of services to the Web, the need for characterizing the users in this environment increases. For that, components that can continually perceive their environment and rapidly adapt to its changes are required, thus reflecting the changes in the user behaviors. Among the existent approaches for modeling Web users, Web usage mining appears as one of the most interesting. Through Web usage mining a user can be implicitly and automatically modeled through his/her usage data, generated from his/her interaction with the Web. This work proposes ONTOWUM, a user modeling and usage mining-based multi-agent framework for Web personalization. The framework comprises two layers, where four agents are distributed according to their responsibilities: Interface agent, responsible for both capturing the user browsing information and performing the adaptation effects; User Modeling agent, responsible for creating and updating both user models and adaptation models; Acquirer agent, responsible for creating and updating an usage data repository containing past user models; and finally, Miner agent, responsible for both discovering group of users with similar browsing behavior and classifying the current user in these groups. The Interface and User modeling agents belong to the user information-processing layer, and the Acquirer and Miner agents belong to the pattern-discovering layer. The methodology and tool used to guide the analyses and design phases were MADEM ( Multi-Agent Domain Engineering Methodology ) and ONTOMADEM, respectively. For the implementation of ONTOWUM it was used the JADE framework.
Com a incessante migração das mais diversas categorias de serviços ao ambiente Web, a necessidade de caracterizar os usuários nesse ambiente nunca foi tão presente. Para isso, são necessários componentes que tenham a habilidade de continuamente perceber o ambiente e rapidamente se adaptarem a ele, refletindo assim as próprias mudanças no comportamento do usuário. Dentre as abordagens existentes para a modelagem de usuários da Web, a mineração de uso da Web figura entre as mais interessantes, pois através dessa abordagem pode-se modelar o usuário implicitamente através dos dados de uso gerados através da sua interação com a Web. Este trabalho propõe ONTOMUW", um framework multiagente para a personalização da Web baseado na modelagem de usuários e na mineração de uso. O framework é composto por duas camadas onde se distribuem quatro agentes de acordo com suas responsabilidades: agente Interfaceador, responsável tanto pela captura das informações provenientes da navegação do usuário quanto pela execução da adaptação; agente Modelador, responsável por criar e atualizar tanto modelos de usuários quanto modelos de adaptação; agente Aquisitor, responsável pela criação e manutenção de um repositório de dados de uso contendo os modelos de usuários passados; e, finalmente, agente Minerador, responsável tanto pela descoberta de grupos de usuários com comportamento de navegação similar quanto pela classificação do usuário corrente nos grupos descobertos. Os agentes Interfaceador e Modelador compõem a camada de processamento de informações do usuário, enquanto que os agentes Aquisitor e Minerador compõem a camada de descoberta de padrões. A metodologia e ferramenta utilizadas para guiar o processo de análise e projeto de domínio do ONTOMUW foram a MADEM ( Multi-Agent Domain Engineering Methodology ) e a ONTOMADEM respectivamente. Para a implementação do framework utilizou-se o ambiente JADE.

Cette thèse analyse le rôle de l’apprentissage dans deux cadres de modélisation distincts. Dans le cas dunouveau modèle canonique avec apprentissage adaptatif, les caractéristiques les plus marquantes des dynamiquesd’apprentissage concernent la capacité des règles de politique monétaire à assurer la convergencevers l’équilibre en anticipations rationnelles. Le mécanisme de transmission de la politique monétaire estcelui de l’effet de substitution associé au canal de la consommation. Dans le cas d’un modèle multi-agentsqui relâche des hypothèses restrictives du nouveau modèle canonique, tout en restant structurellementproche de celui-ci, les variables agrégées évoluent à bonne distance de cet équilibre, et on observe desdynamiques nettement différentes. La politique monétaire influence les variables agrégées de manièremarginale via l’effet de revenu du canal de la consommation. En présence d’un processus d’apprentissagesocial évolutionnaire, l’économie converge vers un faible niveau d’activité économique. L’introductiond’un processus caractérisé par le fait que les agents apprennent individuellement à l’aide de leurs modèlesmentaux atténue le caractère dépressif des dynamiques d’apprentissage. Ces différences entre les deuxcadres de modélisation démontrent la difficulté de généraliser les résultats du nouveau modèle canonique
This thesis analyses the role of learning in two different modelling frameworks. In the new canonicalmodel with adaptive learning, the most remarkable characteristics of the learning dynamics deal withthe capacity of monetary policy rules to guaranty convergence to the rational expectations equilibrium.The transmission mechanism of the monetary policy is based on the substitution effect associated to theconsumption channel. In the case of an agent-based model which relaxes some restrictive assumptionsof the new canonical model - but is endowed with a similar structure - aggregate variables evolve atsome distance from the rational expectations equilibrium. Monetary policy has a marginal impact onthe agregated variables via the wealth effect of the consumption channel. When agents learn accordingto an evolutionnary social learning process, the economy converges to regions of low economic activity.The introduction of a process where agents learn individually by using their mental models induces lessdepressive learning dynamics. These differences between the two modelling frameworks show that thegeneralisation of the results of the new canonical model is not easy to achieve

碩士
國立雲林科技大學
資訊管理系碩士班
92
How to attract the customer to visit as well as how to promote user''s degree of satisfaction is the website gain competitive advantage main keys.The recommendation system can be said a method which can solve, but the present Web Usage Mining Recommendation, all must aim at the website the overall content to design a system to conform to the oneself website recommendation system, but if this system must be used in other websites, then the system must change formula suitability to make this system to be suitable other systems, and provides the customer the most recent information to be possible to say is the count for much, but at present certainly has not made the immediate renewal in view of the entire website material to cause to be unable to recommend the recent information to give the customer. In this study, we proposed a new online incremental web usage mining recommend agent’s methodology. This recommend agent uses incremental mining methodology and uses the website directory structure property to solve the question that as the website data change, the recommendation system also must be able to update. Because this recommend agent uses website directory structure, the agent can maintain the entire recommendation structure immediately, and enables this recommendation system prompt response and renewal recommendation homepage content, also because uses website directory recommendation method therefore the recommendation system can be suitable for the major part website. For three weeks of the experiment, the evaluation result show that: In the perspective of system executing performance, the browsing recommendation ratio was 91.36%, and the average of overall browsing recommendation time was 4.78 seconds, and the average of overall browsing mining time was 18.46 minutes. In the perspective of system recommendation benefit, the overall recommendation precision was 49.5%,and the overall recommendation coverage was 56.7%,and the overall F1 was 52.8%.

碩士
國立雲林科技大學
資訊管理系碩士班
93
To be an outstanding web site, the satisfaction of users is critical. Recommending system is one of the solutions of raising user satisfaction. The mechanisms commonly used by mining recommend system at present are association rules, clustering, classification, and sequence patterns. These technologies are aim to the relationship among web pages, but they omit the time factor. That mentioned above will result in web pages that used to be popular being recommended even if they are out-of-date, and therefore lack of recommendation novelty. In this research, we improve incremental web usage mining recommending system by applying time consideration. We modified the “second chance” method of virtual memory page replacing algorithms to suit recommending system. By giving chances to web pages, we can take novelty and user interests into account simultaneously. We implemented a recommend system and conducted a four-week experiment. In the perspective of system performance, the browsing recommendation ratio was 93.84%, the average of overall browsing recommendation time was 0.059 seconds, and the average of overall browsing mining time was 33.162 minutes. In the perspective of system recommendation benefit, the overall recommendation precision was 43.1%, the overall recommendation coverage was 50.9%, and the overall F1 was 46.7%.

The Internet and world wide web (WWW) have in recent years, grown exponentially in size and in terms of the volume of information that is available on it. In order to effectively deal with the huge amount of information on the web, so called web search engines have been developed for the task of retrieving useful and relevant information for its users. Unfortunately, these web search engines have not kept pace with the boom growth and commercialization of the web. The main goal of this dissertation is the development of a model for a collaborative personalized meta-search agent (COPEMSA) system for the WWW. This model will enable the personalization of web search for users. Furthermore, the model aims to leverage on current search engines on the web as well as enable collaboration between users of the search system for the purposes of sharing useful resources between them. The model also employs the use of multiple intelligent agents and web content mining techniques. This enables the model to autonomously retrieve useful information for it’s user(s) and present this information in an effective manner. In order to achieve the above stated, the COPEMSA model employs the use of multiple intelligent agents. COPEMSA consists of five core components: a user agent, a query agent, a community agent, a content mining agent and a directed web spider. The user agent learns about the user in order to introduce personal preference into user queries. The query agent is a scaled down meta-search engine with the task of submitting the personalized queries it receives from the user agent to multiple search services on theWWW. The community agent enables the search system to communicate and leverage on the search experiences of a community of searchers. The content mining agent is responsible for analysis of the retrieved results from theWWWand the presentation of these results to the system user. Finally, a directed web spider is used by the content mining agent to retrieve the actual web pages it analyzes from the WWW. In this dissertation an additional model is also presented to deal with a specific problem all web spidering software must deal with namely content and link encapsulation.
Prof. E.M. Ehlers

碩士
國立臺灣師範大學
設計研究所
95
In the face of the globalization and market trend, how to extract effective design alternatives from market information and to present a contemporary design process and information system is an important subject. The research objective is to study the statement model and remote design information integration, and then apply the research results to creation. In the process of product development and selling, issues like brand image, market position and design style will be discussed. Meanwhile, consumer behavior is often the result of personal values and life style. In the design process the definition of target market and target consumers’ needs must be ascertained. This study looks into both marketing research and life style and collects relevant information by literature review, interview, and field study. After that, the study creates a statement procedure of remote design information and a fundamental framework of information platform. While learning consumers’ demands for their life from all aspects, designers can recognize the features and cultures of local consumers’ life style and develop their design strategies in accordance with the target consumer’ characteristics, features and preferences. Through the platform, designers combine the information of remote markets for a collaborative design and for the ratiocination of future trend and focus. This study has an insight into Japanese comics and anime and otaku’s life style to testify the above theories and present the influences Japanese comics and anime has on Taiwanese as the reference of future practice and study.

1 Operational and Investigative actions Abstract Presented thesis deals with the topic of operational investigative actions specified by the Criminal Procedure Code, to which they have been incorporated with effect from 2002. The main aim of this thesis is comprehensive description of the essence and of the domestic applicable legislation of these actions, and also the analysis of possible problematic points and deficiencies. Currently, the mentioned Code in its title nine called procedure prior to initiation a criminal prosecution defines types of these actions in an enumerative way. Those types are pretended transfer, surveillance of the individuals and objects and usage of agent. The thesis itself is divided into six chapters. First chapter is focused on definition of these actions, putting them into context with so called supporting operational investigative actions defined by the Police of the Czech Republic Code, and also on the brief history of their usage in the territory of the Czech Republic. In the second chapter there is a summarization of the general conditions concerning usage of these actions, which apply to all operational investigative actions with few exceptions. Third, fourth and fifth chapter deals with particular operational investigative actions according to the order contained in the...