Two testing conditions (immediate and delayed) were used to test for the presence of imitation and emulation in typically developing and developmentally delayed children, including children with autism spectrum disorder, ranging in age from 2.8-years old and 4.0-years old, in two experiments, Experiment I (n=20), Experiment II (n=30). Using a mixed within-between design, I compared the performance of the two groups across various tasks in two testing conditions and analyzed their performance. The participants were selected because they fit the criteria of 1) being between the age of 2.5 and 4 years of age at the onset of the study, 2) had gross motor and generalized imitation in repertoire, and 3) observational learning was present. The independent variable was the test interval in both experiments across both testing conditions, immediate and delayed. The dependent variables were the unconsequated responses during the test interval (Experiment I and II). The embedded dependent variable in Experiment II was the number of 5s intervals participants interacted with a puzzle box in the free play setting. Responses were defined as imitation (copy the specific actions with point-to-point correspondence), or emulation (bring about the model’s goal by the observer’s own methods and means, no point-to-point correspondence but same end result). In the first experiment I found that although typically developing preschoolers often imitate in the short term, they were more likely to emulate in the long term when not shown again how to use the items. In contrast the participants with autism spectrum disorder were more likely to imitate across both testing conditions. My findings support evidence that typically developing children naturally shift from imitation to emulation and that children are in fact emulators in contrast to research that suggests otherwise. For those children with autism, Experiment I, supports evidence that they are potentially missing a developmental cusp (emulation). Experiment II sought to replicate the findings in Experiment I and differed in that 1) more tasks were added, 2) more participants were used, and 2) a free play observation session was added. The results from Experiment II supported the results from Experiment I, in that, all participants (typically developing and those with autism) were more likely to imitate in the short-term immediate testing condition; however, typically developing children naturally shifted to an emulative response given a delay, whereas, those children with autism continued to emit imitative behaviors given a delay, signifying that children with autism are missing the developmental cusp of emulation. The findings support the notion that emulation is a developmental cusp and that children with autism often are missing this developmental cusp.

Imitation learning refers to an agent's ability to mimic a desired behaviour by learning from observations. The field is rapidly gaining attention due to recent advances in computational and communication capabilities as well as rising demand for intelligent applications. The goal of imitation learning is to describe the desired behaviour by providing demonstrations rather than instructions. This enables agents to learn complex behaviours with general learning methods that require minimal task specific information. However, imitation learning faces many challenges. The objective of this thesis is to advance the state of the art in imitation learning by adopting deep learning methods to address two major challenges of learning from demonstrations. Firstly, representing the demonstrations in a manner that is adequate for learning. We propose novel Convolutional Neural Networks (CNN) based methods to automatically extract feature representations from raw visual demonstrations and learn to replicate the demonstrated behaviour. This alleviates the need for task specific feature extraction and provides a general learning process that is adequate for multiple problems. The second challenge is generalizing a policy over unseen situations in the training demonstrations. This is a common problem because demonstrations typically show the best way to perform a task and don't offer any information about recovering from suboptimal actions. Several methods are investigated to improve the agent's generalization ability based on its initial performance. Our contributions in this area are three fold. Firstly, we propose an active data aggregation method that queries the demonstrator in situations of low confidence. Secondly, we investigate combining learning from demonstrations and reinforcement learning. A deep reward shaping method is proposed that learns a potential reward function from demonstrations. Finally, memory architectures in deep neural networks are investigated to provide context to the agent when taking actions. Using recurrent neural networks addresses the dependency between the state-action sequences taken by the agent. The experiments are conducted in simulated environments on 2D and 3D navigation tasks that are learned from raw visual data, as well as a 2D soccer simulator. The proposed methods are compared to state of the art deep reinforcement learning methods. The results show that deep learning architectures can learn suitable representations from raw visual data and effectively map them to atomic actions. The proposed methods for addressing generalization show improvements over using supervised learning and reinforcement learning alone. The results are thoroughly analysed to identify the benefits of each approach and situations in which it is most suitable.

abstract: Recent advancements in external memory based neural networks have shown promise in solving tasks that require precise storage and retrieval of past information. Re- searchers have applied these models to a wide range of tasks that have algorithmic properties but have not applied these models to real-world robotic tasks. In this thesis, we present memory-augmented neural networks that synthesize robot navigation policies which a) encode long-term temporal dependencies b) make decisions in partially observed environments and c) quantify the uncertainty inherent in the task. We extract information about the temporal structure of a task via imitation learning from human demonstration and evaluate the performance of the models on control policies for a robot navigation task. Experiments are performed in partially observed environments in both simulation and the real world / Dissertation/Thesis / Masters Thesis Computer Science 2018

The purpose of this investigation was to determine the effects of a symbolic modeling procedure upon the social adjustment of socially-withdrawn seventh-grade students. The three hypotheses investigated in this study were the following. I. At the conclusion of the experimental period, the mean number of social interactions will be significantly greater for students in the experimental condition than for students in either the placebo or the control condition. II. At the conclusion of the experimental period, the social adjustment of each student in the classroom, as rated by his first-period teacher, will be significantly greater for students in the experimental condition than for students in either the placebo or the control conditions. III. At the conclusion of the experimental period, the personal social adjustment, as measured by a self-rating scale, for the students in the experimental condition will be significantly greater than for students in either the placebo or the control conditions. The following conclusions are presented as a result of this investigation. 1. The symbolic modeling procedure conducted in this study appears to be effective in increasing the frequency of social interaction of socially-withdrawn seventh-grade students even though the findings of the present study did not quite reach the desired level of statistical significance, 2. The symbolic modeling procedure conducted in this study does not appear to be effective with respect to altering the classroom behavior of socially-withdrawn seventh grade students. 3. The symbolic modeling procedure conducted in this study appears to be ineffective with respect to helping seventh-grade students achieve greater personal social adjustment.

Após mais de duzentos anos, o interesse pela obra de Karl Philipp Moritz (15 de setembro de 1756 26 de junho de 1793) só tem aumentado. Diferentes autores como Herman Hesse e Walter Benjamin, e, mais recentemente, Hans Joachin Schrimpf, Tzvetan Todorov, Peter Szondi, Arno Schmidt e Peter Handke têm escrito ressaltando a importância e a fecundidade desse autor. Moritz pode ser considerado um dos autores inaugurais do romantismo alemão. Este mestrado em filosofia, área de estética, pretende, por meio de seleção, tradução e introdução dos textos de Karl Philipp Moritz, contribuir para a valorização dessa importante obra em nossa cultura. Os textos selecionados são de teoria da linguagem, estética e filosofia. / After over two hundred years, the concern for the works of Karl Philipp Moritz (September 15th 1756 June 26th 1793) has increased steadily. Different authors such as Herman Hesse and Walter Benjamin and more recently Hans Joachin Schrimpf, Tzvetan Todorov, Peter Szondi, Arno Schmidt and Peter Handke have written on the relevance and fecundity of this author. Moritz can be said to be one of the inaugural authors of German Romanticism. This Masters in Philosophy, in the Aesthetics field, intends, by means of selection, translation and introduction of Karl Philipp Moritz texts, to contribute to the appreciation of this important work in our culture. The selected texts belong to the fields of Language Theory, Aesthetics and Philosophy.

No description available.

No description available.

The purpose of this study was to test the effects of graduated guidance procedure on teaching imitation of manual signs to students with moderate to severe disabilities. Sessions began with student initiation and were embedded across already established reinforcement routines across the student’s day. A multiple baseline across participants design was used to evaluate these effects.

This study examined, within a multiple baseline design, the development of generalized imitative repertoires and the occurrence of contingency testing in children with autism. Generalized imitation refers to an imitative response class maintained by a conditioned reinforcer: similarity. In the case of imitation, a response class is established by reinforcement of sufficient exemplars of different imitations; the results is a general repertoire of imitating novel responses on their first presentation. Generalized imitation was facilitated through reinforcement procedures involving shaping and fading. Multiple experimenters and multiple settings have been involved to promote the generalization of imitation. Contingency testing, which has been noticed to naturally occur in typically developing children when imitated, was targeted in the examined treatment, by employing a procedure where the experimenter imitated the child. It was hypothesized that the imitation training leads to generalized imitation and that the imitation of the child by the therapist leads to the occurrence of contingency testing. The results of the study show that the shaping and reinforcement procedures determined an increase in participants' trained imitative responding but no corresponding increase in novel, probed imitative responses. The data indicate the emergence of an imitative repertoire in need of further training. Additionally, the results suggest that generalization of skills across therapists and settings has occurred. Imitation of the child by the therapist had been employed for a very small number of sessions. Notwithstanding, the results yielded an increase in the contingency testing behaviors for both participants.

Cette thèse est constituée de quatre essais en économie industrielle. Les deux premiers essais relèvent de l'économie de l'innovation et de la propriété intellectuelle. Ils exploitent l'idée qu'un brevet est un titre incertain et cherchent à en tirer des conséquences, d'une part, sur le choix entre brevet et secret industriel comme régime de protection de l'innovation et, d'autre part, sur les accords de licence entre l'innovateur et les entreprises souhaitant utiliser la nouvelle technologie. Le troisième essai analyse un programme de clémence utilisé par les autorités de la concurrence américaines et visant à inciter des entreprises condamnées pour leur participation à un cartel sur un marché donné à dénoncer d'éventuels comportements collusifs sur d'autres marchés. Le quatrième essai, enfin, s'intéresse à l'effet de la contrainte imposée aux entreprises dans le cadre d'un schéma de régulation par comparaison sur leurs incitations à réduire leurs coûts de production.