No description available.

Yes / Optimization plays a pivotal role in computer graphics and vision. Learning-based optimization algorithms have emerged as a powerful optimization technique for solving problems with robustness and accuracy because it learns gradients from data without calculating the Jacobian and Hessian matrices. The key aspect of the algorithms is the least-squares method, which formulates a general parametrized model of unconstrained optimizations and makes a residual vector approach to zeros to approximate a solution. The method may suffer from undesirable local optima for many applications, especially for point cloud registration, where each element of transformation vectors has a different impact on registration. In this paper, Reweighted Discriminative Optimization (RDO) method is proposed. By assigning different weights to components of the parameter vector, RDO explores the impact of each component and the asymmetrical contributions of the components on fitting results. The weights of parameter vectors are adjusted according to the characteristics of the mean square error of fitting results over the parameter vector space at per iteration. Theoretical analysis for the convergence of RDO is provided, and the benefits of RDO are demonstrated with tasks of 3D point cloud registrations and multi-views stitching. The experimental results show that RDO outperforms state-of-the-art registration methods in terms of accuracy and robustness to perturbations and achieves further improvement than non-weighting learning-based optimization.

Creativity can be a useful skill in today's classrooms and workplaces. When individuals talk about creativity, it's unclear what the controlling variables are when we tact behavior as "creative." Research in understanding the processes behind behaviors that are considered "creative" would assist in identifying functional relations and provide insight on how to teach creativity. Since creativity is often described as doing something different from the norm, behavioral variability may be a potential aspect of creativity. This study aimed to replicate previous findings by investigating the effects of discrimination training in a multiple schedule of varied and repetitive responding in the context of a video game. Participants played through a 2D online video game made in Bloxels. Different alternating-colored platforms served as the discriminative stimuli for the vary and repeat components. Three parameters of variability were measured (e.g., left jumps, right jumps, and double jumps). The results of the study indicate that participants were able to learn the discrimination of when to repeat and vary their responses depending on which colored platform they encountered.

Cette thèse a pour objet l'analyse statistique de formes, dans le contexte de l'imagerie médicale.Dans le champ de l'imagerie médicale, l'analyse de formes est utilisée pour décrire la variabilité morphologique de divers organes et tissus. Nous nous focalisons dans cette thèse sur la construction d'un modèle génératif et discriminatif, compact et non-linéaire, adapté à la représentation de formes.Ce modèle est évalué dans le contexte de l'étude d'une population de patients atteints de la maladie d'Alzheimer et d'une population de sujets contrôles sains. Notre intérêt principal ici est l'utilisationdu modèle discriminatif pour découvrir les différences morphologiques les plus discriminatives entre une classe de formes donnée et des formes n'appartenant pas à cette classe. L'innovation théorique apportée par notre modèle réside en deux points principaux : premièrement, nous proposons un outil pour extraire la différence discriminative dans le cadre Support Vector Data Description (SVDD) ; deuxièmement, toutes les reconstructions générées sont anatomiquementcorrectes. Ce dernier point est dû au caractère non-linéaire et compact du modèle, lié à l'hypothèse que les données (les formes) se trouvent sur une variété non-linéaire de dimension faible. Une application de notre modèle à des données médicales réelles montre des résultats cohérents avec les connaissances médicales. / This thesis addresses statistical shape analysis, in the context of medical imaging. In the field of medical imaging, shape analysis is used to describe the morphological variability of various organs and tissues. Our focus in this thesis is on the construction of a generative and discriminative, compact and non-linear model, suitable to the representation of shapes. This model is evaluated in the context of the study of a population of Alzheimer's disease patients and a population of healthy controls. Our principal interest here is using the discriminative model to discover morphological differences that are the most characteristic and discriminate best between a given shape class and forms not belonging in that class. The theoretical innovation of our work lies in two principal points first, we propose a tool to extract discriminative difference in the context of the Support Vector Data description (SVDD) framework ; second, all generated reconstructions are anatomicallycorrect. This latter point is due to the non-linear and compact character of the model, related to the hypothesis that the data (the shapes) lie on a low-dimensional, non-linear manifold. The application of our model on real medical data shows results coherent with well-known findings in related research.

The purpose of this experiment was to compare the effects of combining negative reinforcement and positive reinforcement during teaching with the effects of using positive reinforcement alone. A behavior was trained under two stimulus conditions and procedures. One method involved presenting the cue ven and reinforcing successive approximations to the target behavior. The other method involved presenting the cue punir, physically prompting the target behavior by pulling the leash, and delivering a reinforcer. Three other behaviors were trained using the two cues contingent on their occurrence. The results suggest that stimuli associated with both a positive reinforcer and an aversive stimulus produce a different dynamic than a situation that uses positive reinforcement or punishment alone.

L’addiction aux drogues peut être considérée comme une maladie neurologique chronique avec des rechutes récurrentes en période d’abstinence qui constituent le problème majeur dans le traitement de l’addiction aux drogues. Grâce à un modèle animal de rechute, il a été montré qu’un rongeur pouvait réinstaller un comportement de recherche de drogue lorsqu’il était réexposé à la drogue elle-même, à des indices associés à la drogue, ou encore à un stress. Dans notre équipe, nous avons évalué la contribution relative des différentes propriétés de la cocaïne, de l’héroïne et de la nicotine (incitative, discriminative et renforçante) dans la réinstallation d’un comportement de recherche de nourriture. Afin de dissocier les propriétés discriminatives et renforçantes, les rats ont été entraînés à s’auto-administrer une récompense alimentaire. Nous avons alors trouvé que : 1) La cocaïne et la nicotine agissent comme des stimuli internes qui acquièrent un contrôle discriminatif sur le comportement, étant donné que la cocaïne et la nicotine, contrairement à l’héroïne, peuvent réinstaller un comportement éteint de recherche de nourriture lorsque ce comportement a été préalablement acquis sous les effets de la cocaïne et de la nicotine, respectivement. 2) La réinstallation induite par la cocaïne et la nicotine est indépendante de la valeur actuelle de la récompense, ce qui indique que la cocaïne et la nicotine contrôlent l’activation de comportements automatiques, habituels, liés à la drogue. Puis, afin d’identifier la façon dont les drogues d’abus entraînent la formation d’habitudes, nous avons également étudié les effets d’une sensibilisation à la cocaïne à différents moments d’un apprentissage instrumental pour une récompense alimentaire, après une dévaluation de la récompense. Nous avons alors trouvé que la sensibilisation à la cocaïne ne favorisait pas le développement de comportements de type habituel. Cette série d’expériences constitue une première étape dans la comparaison des processus automatiques produits par la cocaïne et la nicotine. Si l’activation de comportements de type habituel, automatique, peut être généralisée à d’autres drogues d’abus, nous pourrons considérer que la rechute vers la recherche et la prise de drogue est en partie sous le contrôle de processus automatiques, ce qui pourrait expliquer la forte probabilité de rechute, même après de longues périodes d’abstinence et malgré la connaissance des conséquences néfastes qui en découlent. / Drug addiction can be considered as a chronic brain disease with recurrent relapse during abstinence periods which remains the major problem for the treatment of drug addiction. Using an animal model of drug relapse, it has been shown that a rodent can reinstate a drug-seeking behavior when re-exposed to the drug itself, drug associated cues or stress. In our research group, we assessed the relative contribution of the different properties of cocaine, heroin and nicotine (incentive, discriminative and reinforcing) in food-seeking reinstatement, and in order to dissociate the discriminative from the reinforcing properties, rats were trained to self-administer a non-drug reward (food). We found that: 1) Cocaine and nicotine act as internal stimuli that acquires discriminative control over behavior, since cocaine and nicotine, but not heroin, can reinstate an extinguished food-seeking behavior when this behavior has been previously performed under the effects of cocaine and nicotine respectively. 2) Cocaine- and nicotine-induced reinstatement is independent of the current value of the outcome, which indicates that cocaine and nicotine control the activation of automatic, drug-related habitual behaviors. Then, in order to identify the way drugs of abuse lead to the formation of habits, we also examined the effects of cocaine sensitization at different stages of instrumental training for a food reward after outcome devaluation. We found that, globally, cocaine sensitization does not promote the development of habit-based behaviors. This series of experiments represent a first step in the comparison of automatic processes produced by cocaine and nicotine. If the activation of automatic, habit-based behaviors can be generalized to other drugs of abuse, we could consider that relapse to drug-seeking and drug-taking is partly under the control of automatic processes, which could explain the high probability of relapse, even after extended periods of abstinence and despite the knowledge of the adverse consequences.

Résumé indisponible. / Résumé indisponible.

Made available in DSpace on 2016-04-29T13:18:18Z (GMT). No. of bitstreams: 1 Mariana Cavalcante Vieira.pdf: 1576858 bytes, checksum: a83e1d1db29f1a44bbcb3e8d8910ab78 (MD5) Previous issue date: 2010-04-19 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Metacontingency is the unit of analysis at the cultural level proposed in analogy to the three-term contingency to explain social phenomena that involves the functional relation between interlocking behavioral contingencies (IBC) and its aggregate outcome and the cultural consequence. So far, experimental research has investigated the relation equivalent to the response-reinforcer in the operant contingency. In this study, the following questions were made: (1) An antecedent stimulus similar to the DS assume evocative function to IBC and their aggregate outcome in a metacontingency? (2) What are the effects of alternating two antecedent stimuli, each one correlated to a specific metacontingency? (3) Will the establishment of an analogous to discriminative stimuli control produce an analog of generalization process? Fifteen college students participated in this study, in a total of 12 generations (3 participants in each). This study had seven experimental conditions in an trial procedure. One to three participants worked simultaneously, each one in a computer: the participant of the left (PL), center (PC) and right (PR). In each trial, the computer screen of each participant, presented four numbers, independently, in each column. The participant had to enter another four in matching columns. If the sums of the numbers presented by the computers and the numbers in each column selected by the participant resulted in odd numbers, the participant earned points. This contingency was called individual contingency. Two other contingencies focused on relations between the products of the behavior of the participants, called metacontingencies. In metacontingency 1, the background color of the screen was blue (SM1) and the sum of 4 numbers entered by the participant PL was smaller than the sum of 4 numbers entered by the participant PC and this sum was smaller than the sum of 4 numbers entered by PR (ΣPL <ΣPC <ΣPR), a cultural consequence was produced: the participants received additional credits called bonus. In metacontingency 2, the screens had a red background (SM2) and the participants produced bonus if "ΣPL> ΣPC> ΣPR . These metacontingencies were presented in random order between trials. When stability criteria was reached, older participants were replaced by newer participants. Tests of stimulus control and generalization were presented at established moments. The results showed the selection of operant behavior and metacontingencies 1 and 2 and indicated that SM1 and SM2 acquired evocative function of the corresponding metacontingencies. The tests suggested that the background colors of the screen became the stimulus dimension that exerted control over the behavior of the participants and their interactions. The data are discussed in terms of an analogy between operant contingency and metacontingency / A metacontingência é a unidade de análise no nível cultural proposta em analogia à tríplice contingência para explicar fenômenos sociais que envolvem a relação funcional entre contingências comportamentais entrelaçadas (CCE) e seu produto agregado e uma conseqüência cultural. Até o momento, as pesquisas experimentais investigaram o equivalente à relação resposta-reforçador da contingência operante. No presente estudo, as seguintes perguntas foram feitas: (1) Uma condição de estímulo antecedente análoga ao SD assumiria função evocativa sobre CCEs e seu produto agregado em uma metacontingência? (2) Quais seriam os efeitos de alternar duas condições de estímulo antecedentes, sendo cada uma delas correlacionada a uma metacontingência específica? (3) O estabelecimento do controle de estímulos análogo ao discriminativo produziria processos análogos ao de generalização? Participaram do estudo 15 estudantes universitários, totalizando 12 gerações (3 participantes em cada). O estudo teve sete condições experimentais em procedimento de tentativas. Um a três participantes trabalhavam simultaneamente, cada um em um computador: o participante da esquerda (PE) do centro (PC) e da direita (PD). Em cada tentativa, nas telas dos computadores de cada participante, eram apresentados, independentemente, quatro números um em cada coluna e cabia ao participante inserir outros quatro também em quatro colunas. Se as somas dos números apresentados pelos computadores e dos números selecionados em cada coluna pelo participante resultassem em números ímpares, o participante ganhava pontos. Esta contingência foi chamada de contingência individual. Outras duas contingências incidiam sobre relações entre os produtos dos comportamentos dos participantes, chamadas de metacontingências. Na metacontingência 1, a cor de fundo da tela era azul (SM1) e se a soma dos 4 números inseridos pelo participante PE fosse menor que a soma dos 4 números inseridos pelo participante PC e esta soma fosse menor que a soma dos 4 números inseridos por PD ( ∑PE ∑PC ∑PD ), uma conseqüência cultural era produzida: os participantes recebiam créditos adicionais chamado bônus. Na metacontingência 2, as telas tinham fundo vermelho (SM2) e os participantes produziam bônus se ∑PE ∑PC ∑PD . Estas metacontingências eram apresentadas em ordem aleatória entre tentativas. Atingidos critérios de estabilidade, os participantes antigos eram substituídos por participantes novatos. Em momentos pré-estabelecidos foram conduzidos testes de controle de estímulos e de generalização. Os resultados mostraram a seleção do comportamento operante e das/ pelas metacontingências 1 e 2 e indicaram que SM1 e SM2 adquiriram função evocativa sobre as metacontingências correspondentes. Os testes sugeriram que as cores de fundo da tela tornaram-se a dimensão do estímulo que exerceu controle sobre os comportamentos dos participantes e suas interações. Os dados são discutidos em termos de uma analogia entre contingência operante e metacontingência

Multi-spectral Satellite imagery, among remotely sensed data from airborne and spaceborne platforms, contained the NIR band information is the major source for the land- cover classification. The main purpose of aerial imagery is for thematic land-use/land-cover mapping which is rarely used for land cover classification. Recently, the newly developed digital aerial cameras containing NIR band with up to 10cm ultra high resolution makes the land-cover classification using aerial imagery possible. However, because the urban ground objects are so complex, multi-spectral imagery is still not sufficient for urban classification. Problems include the difficulty in discriminating between trees and grass, the misclassification of buildings due to diverse roof compositions and shadow effects, and the misclassification of cars on roads. Recently, aerial LiDAR (ULiUght UDUetection UAUnd URUanging) data have been integrated with remotely sensed data to obtain better classification results. The LiDAR-derived normalized digital surface models (nDSMs) calculated by subtracting digital elevation models (DEMs) from digital surface models (DSMs) becomes an important factor for urban classification. This study proposed an adaptive raw-data-based, surface-based LiDAR data-filtering algorithm to generate DEMs as the foundation of generating the nDSMs. According to the experiment results, the proposed adaptive LiDAR data-filtering algorithm not only successfully filters out ground objects in urban, forest, and mixed land cover areas but also derives DEMs within the LiDAR data measuring accuracy based on the absolute and relative accuracy evaluation experiments results. For the aerial imagery urban classification, this study first conducted maximum likelihood classification (MLC) experiments to identify features suitable for urban classification using LiDAR data and aerial imagery. The addition of LiDAR height data improved the overall accuracy by up to 28 and 18%, respectively, compared to cases with only red¡Vgreen¡Vblue (RGB) and multi-spectral imagery. It concludes that the urban classification is highly dependent on LiDAR height rather than on NIR imagery. To further improve classification, this study proposes a knowledge-based classification system (KBCS) that includes a three-level height, ¡§asphalt road, vegetation, and non-vegetation¡¨ (A¡VV¡VN) classification model, rule-based scheme and knowledge-based correction (KBC). The proposed KBCS improved overall accuracy by 12 and 7% compared to maximum likelihood and object-based classification, respectively. The classification results have superior visual interpretability compared to the MLC classified image. Moreover, the visual details in the KBCS are superior to those of the OBC without involving a selection procedure for optimal segmentation parameters.

This dissertation presents the development of a semi-supervised incremental learning framework with a multi-view perspective for image concept modeling. For reliable image concept characterization, having a large number of labeled images is crucial. However, the size of the training set is often limited due to the cost required for generating concept labels associated with objects in a large quantity of images. To address this issue, in this research, we propose to incrementally incorporate unlabeled samples into a learning process to enhance concept models originally learned with a small number of labeled samples. To tackle the sub-optimality problem of conventional techniques, the proposed incremental learning framework selects unlabeled samples based on an expected error reduction function that measures contributions of the unlabeled samples based on their ability to increase the modeling accuracy. To improve the convergence property of the proposed incremental learning framework, we further propose a multi-view learning approach that makes use of multiple features such as color, texture, etc., of images when including unlabeled samples. For robustness to mismatches between training and testing conditions, a discriminative learning algorithm, namely a kernelized maximal- figure-of-merit (kMFoM) learning approach is also developed. Combining individual techniques, we conduct a set of experiments on various image concept modeling problems, such as handwritten digit recognition, object recognition, and image spam detection to highlight the effectiveness of the proposed framework.