The World-Wide Web contains a lot of information and reading through the web pages to collect this information is tedious, time consuming and error prone. Users need an automated solution for extracting or highlighting the data that they are interested in. Building a regular expression to match the text they are interested in will automate the process, but regular expressions are hard to create and certainly are not feasible for non-programmers to construct. Text Identification by Example (TIBE) makes it easier for end-users to harvest information from the web and other text documents. With TIBE, training text classifiers from user-selected positive and negative examples replaces the hand-writing of regular expressions. The text classifiers can then be used to extract or highlight text on web pages.

Self-driving cars rely on their sense of sight to function effectively in chaotic and uncontrolled environments. Thanks to recent developments in computer vision, specifically convolutional neural networks, autonomous vehicles have developed the ability to see at or above human-level capabilities, which in turn has allowed for rapid advances in self-driving cars. Unfortunately, much like humans being confused by simple optical illusions, convolutional neural networks are susceptible to simple adversarial inputs. As there is no overlap between the optical illusions that fool humans and the adversarial examples that threaten convolutional neural networks, little is understood as to why these adversarial examples dupe such advanced models and what effective mitigation techniques might exist to resolve these issues. This thesis focuses on these adversarial images. By extending existing work, this thesis is able to offer a unique perspective on adversarial examples. Furthermore, these extensions are used to develop a novel attack that can generate physically robust adversarial examples. These physically robust instances provide a unique challenge as they transcend both individual models and the digital domain, thereby posing a significant threat to the efficacy of convolutional neural networks and their dependent applications.

No description available.

No description available.

In this report a discriminative and a generative image classifier, used for classification of images with handwritten digits from zero to nine, are compared. The aim of this project was to compare the accuracy of the two classifiers in absence and presence of perturbations to the images. This report describes the architectures and training of the classifiers using PyTorch. Images were perturbed in four ways for the comparison. The first perturbation was a model-specific attack that perturbed images to maximize likelihood of misclassification. The other three image perturbations changed pixels in a stochastic fashion. Furthermore, The influence of training using perturbed images on the robustness of the classifier, against image perturbations, was studied. The conclusions drawn in this report was that the accuracy of the two classifiers on unperturbed images was similar and the generative classifier was more robust against the model-specific attack. Also, the discriminative classifier was more robust against the stochastic noise and was significantly more robust against image perturbations when trained on perturbed images. / I den här rapporten jämförs en diskriminativ och en generativ bildklassificerare, som används för klassificering av bilder med handskrivna siffror från noll till nio. Syftet med detta projekt var att jämföra träffsäkerheten hos de två klassificerarna med och utan störningar i bilderna. Denna rapport beskriver arkitekturerna och träningen av klassificerarna med hjälp av PyTorch. Bilder förvrängdes på fyra sätt för jämförelsen. Den första bildförvrängningen var en modellspecifik attack som förvrängde bilder för att maximera sannolikheten för felklassificering. De andra tre bildförvrängningarna ändrade pixlar på ett stokastiskt sätt. Dessutom studerades inverkan av träning med störda bilder på klassificerarens robusthet mot bildstörningar. Slutsatserna som drogs i denna rapport är att träffsäkerheten hos de två klassificerarna på oförvrängda bilder var likartad och att den generativa klassificeraren var mer robust mot den modellspecifika attacken. Dessutom var den diskriminativa klassificeraren mer robust mot slumpmässiga bildförvrängningar och var betydligt mer robust mot bildstörningar när den tränades på förvrängda bilder. / Kandidatexjobb i elektroteknik 2022, KTH, Stockholm

No description available.

Mathematics remains a subject many students fail to become competent in by the time they graduate from high school. Most students often require one on one, individualized tutoring to help them reach competence. That remains a challenge since most classrooms are understaffed and underfunded, frequently having only one teacher in a overpopulated classroom. One strategy that has been used to alleviate some of this over reliance on the teacher has been faded worked examples, or fading. Fading is the successive removal of the last steps in a series of problems until the student is solving problems completely on their own. The current study aimed to determine whether fading improves learning, and for whom. The goal was to compare fading with business as usual (control), worked examples with self-explanations, and fading with self-explanations. Specifically, I was interested in the following research questions: (1) Do the three experimental conditions differ in promoting posttest scores on surface area and volume? (2) Do the three experimental conditions differ in promoting conceptual knowledge and procedural knowledge of surface area and volume at posttest? and (3) When interaction terms are created between student profiles and conditions within regression analyses, which profiles explain significant variance in posttest scores? Repeated measures analysis of variance, principle axis factor analysis, and simple linear regressions were used to examine the differences between conditions at posttest, to create propensity scores, and to determine whether there were any interactions between propensity scores and conditions. Results indicated a significant effect of fading on posttest scores. A regression with propensity factors indicated that the fading conditions appeared to benefit low propensity students moreso than high propensity students. Findings are discussed in terms of educational implications and future research that can complement these findings to contribute to future research. / Educational Psychology

[pt] Esta dissertação apresenta uma solução que permite a geração automática de exemplos de uso a partir da descrição textual de casos de uso. Os casos de uso descrevem especificações em um nível de formalização suficiente para a geração dos exemplos. Um exemplo gerado é um texto em linguagem natural que é o resultado da paráfrase de um possível comportamento do software, extraído de um caso de uso e aplicado a um contexto real, em que atores são convertidos em personagens fictícios e os atributos são valorados de acordo com as regras de negócios especificadas no caso de uso. O formato proposto para a construção de exemplos tem como objetivo permitir que clientes possam ler, entender e julgar se o comportamento que está sendo proposto é o desejado. Com isso é esperado que o próprio cliente possa validar as especificações e que, quando defeitos forem encontrados, a especificação possa logo ser corrigida e refletida de volta nos exemplos. Ao mesmo tempo a especificação formalizada na forma de um caso de uso auxiliará desenvolvedores a criar soluções mais próximas do correto por construção, quando comparado com especificações textuais convencionais. / [en] This master s dissertation presents a solution for the automatic generation of examples of use from the textual description of use cases. Use cases describe specifications in a sufficiently formal way that is enough to automatically generate usage examples. A generated example is a text in a natural language which is the paraphrase of one possible manner to use the software, extracted from the use case and applied to a real context where actors are converted into fictitious personas and attributes are valued according to the business rules specified in the use case. The proposed format to present the example aims to allow clients to read, to understand and to judge whether the expressed behavior is in fact what he wants. With this approach, it is expected that the customer himself can approve the specifications and when defects are found, so the specification can quickly be corrected and reflected in the examples. At the same time, the formalized specification in the form of a use case will help developers create solutions that are by construction closer to the correct one when compared to conventional textual specifications.

The thesis researches the ability of use of Python programming language in the field of high school education and uses qualitative and quantitative methods of research and finds Python suitable. The thesis also brings theoretical analysis of the Python language, including practical exams in the form of sample programs. It researchs even other programming languages used in education and their suitability for this purpose and compares them with Python programming language. It also finds two most common used programming languages at high schools, which are Pascal / Object Pascal and Java.

Cette thèse s'inscrit dans le cadre du traitement automatique du langage naturel, et traite plus précisément de l'alignement sous-phrastique bilingue classiquement lié à la traduction automatique statistique. Les travaux exposés s'en distinguent en proposant une mécanique évolutive à base d'exemples initiée par des annotateurs non-experts via une interface adaptée. L'approche est principalement motivée par la recherche d'une expressivité comparable à celle observée dans les alignements manuels. Une partie importante de ce travail consiste à définir un cadre formel sous-tendant une architecture originale à base d'exemples alignés. Plusieurs mémoires d'alignements ont été constituées en tirant parti d'informations provenant d'analyseurs syntaxiques automatiques en plaçant les prérequis technologiques à un niveau raisonnablement peu élevé. Deux nouvelles méthodes d'alignement sont comparées à des références connues via des mesures d'accord classiques et trois distances transformationnelles sont introduites. / This research belongs to the Natural Language Processing (NLP) field and more specifically focuses on topic Sub-sentential Alignment which is closely related to Machine Translation. The originality of this work consists in an example-based approach bootstrapped by the participation of non-expert annotators through an appropriate interface. Seeking for a greater expressivity, such as observed in manual alignments, mainly motivates the whole approach. An important effort has been made to define a formal environment for this original architecture based on aligned examples. Several memories have been created using syntactic informations from parsers' outputs with reasonnable low-tech requirements. A couple of new alignment methods were compared with state-of-the-art measures and three transformational metrics were introduced.