Access technologies often deal with the classification of several physiological and biomechanical signals. In most previous studies involving access technologies, a single classifier has been trained. Despite reported success of these single classifiers, classification accuracies are often below clinically viable levels. One approach to improve upon the performance of these classifiers is to utilize the state of- the-art multiple classifier systems (MCS). Because MCS invoke more than one classifier, more information can be exploited from the signals, potentially leading to higher classification performance than that achievable with single classifiers. Moreover, by decreasing the feature space dimensionality of each classifier, the speed of the system can be increased. MCSs may combine classifiers on three levels: abstract, rank, or measurement level. Among them, abstract-level MCSs have been the most widely applied in the literature given the flexibility of the abstract level output, i.e., class labels may be derived from any type of classifier and outputs from multiple classifiers, each designed within a different context, can be easily combined. In this thesis, we develop two new abstract-level MCSs based on "reputation" values of individual classifiers: the static reputation-based algorithm (SRB) and the dynamic reputation-based algorithm (DRB). In SRB, each individual classifier is applied to a “validation set”, which is disjoint from training and test sets, to estimate its reputation value. Then, each individual classifier is assigned a weight proportional to its reputation value. Finally, the total decision of the classification system is computed using Bayes rule. We have applied this method to the problem of dysphagia detection in adults with neurogenic swallowing difficulties. The aim was to discriminate between safe and unsafe swallows. The weighted classification accuracy exceeded 85% and, because of its high sensitivity, the SRB approach was deemed suitable for screening purposes. In the next step of this dissertation, I analyzed the SRB algorithm mathematically and examined its asymptotic behavior. Specifically, I contrasted the SRB performance against that of majority voting, the benchmark abstract-level MCS, in the presence of different types of noise. In the second phase of this thesis, I exploited the idea of the Dirichlet reputation system to develop a new MCS method, the dynamic reputation-based algorithm, which is suitable for the classification of non-stationary signals. In this method, the reputation of each classifier is updated dynamically whenever a new sample is classified. At any point in time, a classifier’s reputation reflects the classifier’s performance on both the validation and the test sets. Therefore, the effect of random high-performance of weak classifiers is appropriately moderated and likewise, the effect of a poorly performing individual classifier is mitigated as its reputation value, and hence overall influence on the final decision is diminished. We applied DRB to the challenging problem of discerning physiological responses from nonverbal youth with severe disabilities. The promising experimental results encourage further development of reputation-based multi-classifier systems in the domain of access technology research.

La thèse présentée dans ce rapport a été entreprise dans le cadre de la convention CIFRE (Convention Industrielle de Formation par la REcherche) 264/85 établie entre la société SYNC et l'Ecole Nationale Supérieure des Mines de Saint-Etienne. La réalisation de cette thèse a comporté quatre activités très différentes, à savoir: - L' acquisition des connaissances nécessaires à l' administration et à l' installation d'UNIX. - l'acquisition des compétences nécessaires au développement d'un système expert. - le choix d'un environnement de développement de systèmes experts adapté au domaine d'expertise. - la réalisation du système expert pour installer UNIX.

Des dels anys 80, els sistemes basats en el coneixement, programes que utilitzen una gran quantitat de informació per modelar situacions i resoldre problemes, han adquirit gran importància en el camp industrial, financer i científic. La complexitat d'aquests sistemes fa que el seu ús presenti més dificultats que altres aplicacions informàtiques. La comunicació entre els sistemes basats en el coneixement i l'usuari presenta, doncs, nous reptes. Tot i que el llenguate natural es especialment apropiat per comunicar-se amb aquests sistemes, són pocs els que incorporen interfícies en llenguatge natural. Els motius principals són els problemes d'eficiència que presenta el processament del llenguatge natural i l'elevat cost de desenvolupar les bases de coneixement (conceptual i lingüístic) necessàries per a cada aplicació. L'objectiu d'aquesta tesi és millorar la comunicació en llenguatge natural amb els sistemes basats en el coneixement. Aquesta recerca s'ha centrat en el disseny d'una representació reutilitzable dels diferents tipus de coneixement involucrats en aquesta comunicació, que permetir de generar de forma automàtica la interfície més adequada per a cada aplicació. S'ha desenvolupat un sistema, GISE (Generador de Interfaces a Sistemas Expertos), que genera interfícies en llenguatge natural per diferents tipus d'aplicacions. Aquest sistema adapta automàticament les bases de coneixement lingüístic generals als requeriments d'una aplicació concreta, obtenint la gramàtica més apropiada. El disseny del sistema està basat en una representació reutilitzable i modular dels diferents tipus de coneixement necessaris en la comunicació en llenguatge natural. Aquesta informació consisteix en els conceptes de l'aplicació, les tasques de comunicació, el coneixement lingüístic i les relacions generals entre el coneixement conceptual i la seva realització lingüística. Tres bases de coneixement s'han dissenyat per representar aquesta informació: la ontologia conceptual, la ontologia lingüística i un conjunt de relges de producció. El coneixement conceptual s'ha representat en la ontologia conceptual. Aquest coneixement inclou aspectes sobre el domini i la funcionalitat. Tota la informació necessària per modelar l'aplicació i tots els possibles actes de comunicació estan representats en la ontologia conceptual. La complexitat dels sistemes basats en el coneixement fa necessària una representació formal i explícita de la seva funcionalitat i domini.El coneixement lingüístic general necessari per expressar en llenguatge natural les possibles tasques del sistema es representen en la ontologia lingüística.La informació que permet relacionar el coneixement lingüístic general a una aplicació concreta per tal d'obtenir la gramàtica més adequada es representada mitjançant un conjunt de regles de producció.L'organització modular dels diferents tipus de coneixement que intervenen en la comunicació facilita l'adaptació del sistema a diferents tipus d'aplicacions i usuaris.Les gramàtiques generades pel sistema GISE utilitzen un llenguatge alhora ric i precís, adaptat a l'aplicació. La interfície del sistema incorpora un sistema de finestres que guia a l'usuari a introduir les opcions en llenguatge natural que el sistema reconeix.GISE s'ha aplicat a diferents sistemes: a SIREDOJ, un sistema expert en lleis i a un sistema que dóna informació sobre trens. / Since the 1980's, knowledge based systems (KBSs), programs that use knowledge to model situations and solve problems, have spread throughout industry, finance and science. Human communication with these systems deals with complex concepts and relationships that are not present in other software applications. Allthough the natural language (NL) is especially appropriate for expressing these concepts, there are not many KBSs incorporating NL interfaces. The main reasons for this are problems of efficiency in NLI performance, lack of adequacy to the communication needs of the applications and the high cost of developing and maintaining them.The aim of this thesis is to study how the communication process and engineering features can be improved in NL interaction with KBSs. This study has been focused on the efficient and reusable representation of the knowledge involved in NL communication with KBSs. GISE (Generador de Interfaces a Sistemas Expertos), a system supporting NL communication with KBSs has been developed. This system adapts the general linguistic resources to application requirements in order to automatically obtain application-restricted grammars. The main issue of the system design is a separate and reusable representation of all types of knowledge involved in communication with KBSs. This knowledge consists of the application knowledge appearing in the communication, the tasks of communication, the linguistic knowledge supporting their expression and the general relationships between conceptual knowledge and its linguistic realization. Three general bases were designed to represent all this knowledge : the Conceptual Ontology (CO), the Linguistic Ontology (LO) and a set of control rules.Conceptual knowledge is represented in the CO. This conceptual knowledge includes domain and functionality issues. All knowledge required to model the applications as well as the description of all possible communication acts is provided in the CO. The CO is the skeleton for anchoring the domain and the functionality of the applications. The complexity of KBS performance makes a formal and explicit representation of their domain and functionality necessary. The general linguistic knowledge needed to cover the expression in NL of the tasks the system performs is represented by means of the LO and a set containing all possible realizations of the application terms. The LO is domain and application independent. The control information to relate the general linguistic knowledge to conceptual application knowledge in order to generate the application-restricted grammars is represented by a set of production rules. The modular organization of the relevant knowledge into separate data structures provides great flexibility for adapting the system to different types of applications and users.The grammars generated by GISE use expressive and precise language tuned to the application and adapted to the evolution of the communicative process. A menu-system to guide the user in introducing the NL is integrated into the GISE interface. GISE has been applied to a couple of applications: SIREDOJ, an ES in law and a railway communication system.

This thesis examines the problem of having an intelligent agent reasoning about interaction with users in real-time decision making environments. Our work is motivated by the models of Fleming and Cheng, which reason about interaction sensitive to both expected quality of decision (following interaction) and cost of bothering users. In particular, we are interested in dynamic, time critical scenarios. This leads first of all to a novel process known as strategy regeneration, whereby the parameter values representing the users and the task at hand are refreshed periodically, in order to make effective decisions about which users to interact with, for the best decision making. We also introduce two new parameters that are modeled: each user's lack of expertise (with the task at hand) and the level of criticality of each task. These factors are then integrated into the process of reasoning about interaction to choose the best overall strategy, deciding which users to ask to resolve the current task. We illustrate the value of our framework for the application of decision making in hospital emergency room scenarios and offer validation of the approach, both through examples and from simulations. To sum up, we provide a framework for reasoning about interaction with users through user modeling for dynamic environments. In addition, we present some insights into how to improve the process of hospital emergency room decision making.

Expert finding is an Information Retrieval (IR) task that is used to find the needed experts. To find the needed experts is a noticeable problem in many commercial, educational or governmental organizations. It is highly crucial to find the appropriate experts, when seeking referees for a paper submitted to a conference or when looking for a consultant for a software project. It is also important to find the similar experts in case of the absence or the inability of the selected expert. Traditional expert finding methods are modeled based on three components which are a supporting document collection, a list of candidate experts and a set of pre-defined topics. In reality, most of the time pre-defined topics are not available. In this study, we propose an expert finding system which generates a semantic layer between domains and experts using Latent Dirichlet Allocation (LDA). A traditional expert finding method (voting approach) is used in order to match the domains and the experts as the baseline method. In case similar experts are needed, the system recommends experts matching the qualities of the selected experts. The proposed model is applied to a semi-synthetic data set to prove the concept and it performs better than the baseline method. The proposed model is also applied to the projects of the Technology and Innovation Funding Programs Directorate (TEYDEB) of The Scientific and Technological Research Council of Turkey (T&Uuml / B

<p>In the recent years a growing interest in Cognitive Science has been directed to the cognitive role of the agent's ability to predict the consequences of their actions, without actual engagement with their environment. The creation of an experimental model for Hesslow's simulation hypothesis, based on the use of a simulated adaptive agent and the methods of evolutionary robotics within the general perspective of radical connectionism, is reported in this dissertation. A hierarchical architecture consisting of a mixture of (recurrent) experts is investigated in order to test its ability to produce an 'inner world', functional stand-in for the agent's interactions with its environment. Such a mock world is expected to be rich enough to sustain 'blind navigation', which means navigation based solely on the agent's own internal predictions. The results exhibit the system's vivid internal dynamics, its critical sensitivity to a high number of parameters and, finally, a discrepancy with the declared goal of blind navigation. However, given the dynamical complexity of the system, further analysis and testing appear necessary.</p>

Thèse doctorat : Informatique : Villeurbanne, INSA : 2005. / Titre provenant de l'écran-titre. Bibliogr. p. 176-182.

Thèse de doctorat : Informatique : Nancy 1 : 2001. / Thèse : 2001NAN10123. Titre provenant de l'écran-titre.

Reproduction de : Thèse doctorat : Informatique : Reims : 2008. / Titre provenant de l'écran titre. Bibliogr. p. 101-106.

Thèse de doctorat : Sciences et Techniques du déchet : Villeurbanne, INSA : 1999. / Contient 1 glossaire. Bibliogr. p. 211-217.