Improvements in both software and curricula have helped introductory computer science courses attract and retain more students. Pythy is one such online learning environment that aims to reduce software setup related barriers to learning Python while providing facilities like course management and grading to instructors. To further enable its goals of being beginner-centric, we want to integrate full support for media-computation-style programming activities. The media computation curriculum teaches fundamental computer science concepts through the activities of manipulating images, sounds and videos, and has been shown to be successful in retaining students and helping them gain transferable knowledge. In this work we tackle the first two installments of the problem namely, supporting image and sound-based media computation programs in Pythy. This involves not only client-side support that enables students to run media-computation exercises in the browser, but also server-side support to leverage Pythy's auto-grading facilities. We evaluated our implementation by systematically going through all 82 programs in the textbook that deal with image and sound manipulation and verifying if they worked in Pythy as-is, while complementing this with unit-tests for full test coverage. As a result, Pythy now supports 65 out of the 66 media-computation methods required for image and sound manipulation on both the client and the server-side, and 81 out of the 82 programs in the media-computation textbook can be executed as-is in Pythy. / Master of Science

SQL injection attacks have been a problem since the early 2000s. Even though the issue is well known today, SQL injection vulnerabilities remain to be one of the most common security flaws. Our study looks at open source-projects written in Java and Python to examine the current state of SQL injection vulnerabilities. We take a deeper look into the vulnerabilities' code patterns and discuss suitable prevention methods.The open-source projects were mined from Github. The projects that contained a connection to a relational database were further analyzed with static analysis to find vulnerabilities. A subset of projects was picked out for deeper analysis of the source code. Among 167,644 Java projects that met the selection criteria, 24,416 were identified to have a connection to a relational database. The corresponding figures for Python were 294,637 and 20,994. Notably, concatenation is more prevalent in Java, whereas prepared statements are favored in Python. Additionally, the analysis revealed that the Python projects tended to have more recent updates and a higher number of contributors compared to the Java projects. Moreover, projects employing both prepared statements and concatenation were observed to be larger in size compared to those using only one of these methods or relying on hardcoded queries. With legacy projects and concatenation being more common in Java, it is suggested that Python projects better follow best practices when it comes to SQL injection. Although indications were found that the overall knowledge of SQL injection has increased since 2019, the use of identifier concatenation still is prevalent.

Le fonctionnement du schéma de surface canadien CLASS a été analysé en détail, basé sur une démarche de rétroconception. L’impact des multiples variables d’états du modèle sur les termes des bilans énergétique et hydrique a été expliqué. La valeur des albédos et des transmissivités de la canopée augmente en fin de saison lorsque la canopée devient moins dense. Donc, le rayonnement au sol augmente alors que celui à la canopée diminue. La résistance de couche limite de la feuille ralentit les transferts de chaleur sensible et latente à la canopée durant le jour, mais n’a aucune influence la nuit. La résistance aérodynamique au transfert de chaleur est plus élevée le jour que la nuit. Elle influe sur les flux de chaleur sensible et latente à la canopée. La résistance de surface au transfert de chaleur est très élevée le jour et peu élevée la nuit. Elle influe sur les flux de chaleur sensible et latente au sol. La résistance stomatale est très grande la nuit. Elle freine le transfert de chaleur latente durant le jour et n’a aucune influence sur les flux de chaleur sensible. Finalement, on a observé de grandes fluctuations de température et de teneur en eau dans les deux premières couches de sol. Tandis que la troisième couche de sol a montré une réaction très lente aux précipitations et aux variations de température à la surface du sol. Les résultats sont supportés d’explications théoriques très détaillées dans la section théorie. / Canadian Land Surface Scheme

The DataFrame is a powerful table-like data structure used frequently in Data Science, the in-demand and innovative field focused on the extraction of valuable insights from data. Typically, datasets are not perfect upon collection and need to be prepared so that the resulting dataset is useful for statistical analysis. A DataFrame API supports optimized methods such as, selecting, aggregating and filtering rows, columns, and cells as well as renaming row and column labels. It also supports methods for normalizing data, merging data, adding new columns and labelling missing data among numerous other features. An API to work with tabular data would be useful in any general purpose language, so DataFrames have been incorporated into libraries like Pandas for Python and provided as native libraries in the languages R and Scala. Due to their wide-ranging use it is not uncommon to find implementations in many other languages like Java and Julia \cite{BigData}. In this work, we introduce RacketFrames, a Racket V8.0+ DataFrame implementation. We show the benefits an implementation can have on existing and future Racket projects. To quantify the performance of major DataFrame operations, we measure speed against Python Pandas and compare functional and object oriented paradigms. We hope to continue the trend for further Data Science tool development for Racket and other programming languages.

This thesis presents a modern approach to the Method of Integral Relations implemented in the Python programming language. This work is based on South's reports on solving supersonic conical flows and Belotserkovskii's publications on solving supersonic flows over blunt bodies. The Python SciPy library is extensively used in this thesis. A root finder module is used to perform South's solution process, and the Runge-Kutta ODE solver with dense output is implemented to replicate Belotserkovskii's solution process. The implementation of Python makes the development efficient and serves as an open-source framework for future work to build on. This work focuses on the 1-strip approximation of MIR, and the results are compared against Belotserkovskii's 2 and 3-strip results, experimental data, and the modified Newtonian method. The current 1-strip method shows the agreements of predicting the normal shock standoff distance and the shock wave shape to results from both Belotserkovskii and experiments. Future work on higher-order approximations and additional high-temperature models is required to improve the prediction of MIR in the hypersonic flow region.

Titre de l'écran-titre (visionné le 29 novembre 2023) / L'apprentissage automatique a considérablement progressé dans le domaine de la recherche en neurosciences, mais son application pose des défis en raison des différences entre les principes biologiques du cerveau et les méthodes traditionnelles d'apprentissage automatique. Dans ce contexte, le projet présenté propose NeuroTorch, un pipeline convivial d'apprentissage automatique spécialement conçu pour les neuroscientifiques, afin de relever ces défis. Les objectifs clés de ce projet sont de fournir une librairie d'apprentissage profond adaptée aux neurosciences computationnelles, d'implémenter l'algorithme eligibility trace forward propagation (e-prop) pour sa plausibilité biologique, de comparer les réseaux de neurones continus et à impulsions en termes de résilience, et d'intégrer un pipeline d'apprentissage par renforcement. Le projet se divise en plusieurs parties. Tout d'abord, la théorie des dynamiques neuronales, des algorithmes d'optimisation et des fonctions de transformation d'espaces sera développée. Ensuite, l'attention sera portée sur la conception du pipeline NeuroTorch, incluant l'implémentation de l'algorithme e-prop. Les résultats de la prédiction de séries temporelles d'activité neuronale chez le poisson-zèbre seront présentés, ainsi que des observations sur la résilience à l'ablation des réseaux obtenus. Enfin, une section sera consacrée à l'exploration du pipeline d'apprentissage par renforcement de NeuroTorch et à la validation de son architecture dans l'environnement LunarLander de Gym. En résumé, les modèles à impulsions de NeuroTorch ont atteint des précisions de 96,37%, 85,58% et 74,16% respectivement sur les ensembles de validation MNIST, Fashion-MNIST et Heidelberg. De plus, les dynamiques leaky-integrate-and-fire with explicit synaptic current - low pass filter (SpyLIF-LPF) et Wilson-Cowan ont été entraînées avec succès à l'aide de l'algorithme e-prop sur des données neuronales expérimentales du ventral habenula du poisson-zèbre, obtenant respectivement des valeurs de pVar de 0,97 et 0,96. Les résultats concernant la résilience indiquent que l'application de la loi de Dale améliore la robustesse des modèles en termes d'ablation hiérarchique. Enfin, grâce au pipeline d'apprentissage par renforcement de NeuroTorch, différents types d'agents inspirés des neurosciences ont atteint le critère de réussite dans l'environnement LunarLander de Gym. Ces résultats soulignent la pertinence et l'efficacité de NeuroTorch pour les applications en neurosciences computationnelles. / Machine learning has made significant advancements in neuroscience research, but its application presents challenges due to the differences between the biological principles of the brain and traditional machine learning methods. In this context, the presented project proposes NeuroTorch, a comprehensive machine learning pipeline specifically designed for neuroscientists to address these challenges. The key objectives of this project are to provide a deep learning library tailored to computational neuroscience, implement the eligibility trace forward propagation (e-prop) algorithm for biological plausibility, compare continuous and spiking neural networks in terms of resilience, and integrate a reinforcement learning pipeline. The project is divided into several parts. Firstly, the theory of neural dynamics, optimization algorithms, and space transformation functions will be developed. Next focus will be on the design of the NeuroTorch pipeline, including the implementation of the e-prop algorithm. Results of predicting a time series of neuronal activity in zebrafish will be presented, along with observations on the resilience to network ablations obtained. Finally, a section will be dedicated to exploring the NeuroTorch reinforcement learning pipeline and validating its architecture in the LunarLander environment of Gym. In summary, NeuroTorch spiking models achieved accuracies of 96.37%, 85.58%, and 74.16% on the MNIST, Fashion-MNIST, and Heidelberg validation sets, respectively. Furthermore, the leaky-integrate-and-fire with explicit synaptic current - low pass filter (SpyLIF-LPF) and Wilson-Cowan dynamics were successfully trained using the e-prop algorithm on experimental neuronal data from the ventral habenula of zebrafish, achieving pVar values of 0.97 and 0.96, respectively. Results regarding resilience indicate that the application of the Dale law improves the robustness of models in terms of hierarchical ablation. Lastly, through the NeuroTorch reinforcement learning pipeline, different types of neuroscience-inspired agents successfully met the success criterion in the Gym's LunarLander environment. These results highlight the relevance and effectiveness of NeuroTorch for applications in computational neuroscience.

Existem várias linguagens e plataformas que permitem a programação baseada no modelo de atores, uma solução elegante para a programação concorrente proposta há algumas décadas. Segundo esse modelo, implementa-se o programa na forma de uma série de agentes que são executados em paralelo e se comunicam entre si somente por meio da troca de mensagens, sem a necessidade de memória compartilhada ou estruturas tradicionais de sincronização como semáforos e mutexes. Uma das áreas nas quais esse modelo seria particularmente adequado é a programação de aplicações web, isto é, aplicações cujas lógicas de negócios e de dados residem num servidor e que são acessadas pelo usuário por intermédio de um navegador. Porém, existem muitos obstáculos ao desenvolvimento de aplicações desse tipo, entre eles a falta de linguagens e ferramentas que permitam integrar tanto o servidor quanto o cliente (navegador) no modelo de atores, as dificuldades de conversões de dados que se fazem necessárias quando o servidor e o cliente são desenvolvidos em linguagens diferentes, e a necessidade de contornar as dificuldades inerentes aos detalhes do protocolo de comunicação (HTTP). O PAWEB é uma proposta de uma plataforma para o desenvolvimento e execução de aplicações web que fornece a infraestrutura necessária para que tanto o lado cliente quanto o lado servidor do aplicativo hospedado possam ser escritos numa mesma linguagem (Python), e possam criar e gerenciar atores que trocam mensagens entre si,tanto local quanto remotamente, de maneira transparente e sem a necessidade de implementar conversões de dados ou outros detalhes de baixo nível. / There are several programming languages and platforms that allow the development of systems based on the actor model, an elegant solution for concurrent programming proposed a few decades ago. According to this model, the program is implemented in the form of several agents that run concurrently and only communicate amongst themselves through the exchange of messages, without the need for shared memory or traditional synchronization structures such as semaphores and mutexes. One of the areas where this model would be particularly appropriate would be the development of web applications, that is, applications whose business and database logic reside on the server and are accessed by the user by means of a web browser. However, there are several obstacles to the development of this type of application, amongst them the lack of languages and tools that allow for the integration of both the server and the client (browser) into the actor model, data conversion difficulties arising from using different programming languages on the server and the client, and the need to circumvent the inherent difficulties posed by the details of the communications protocol (HTTP). PAWEB is a proposal for an application development and execution platform that supplies the infrastructure needed so that both the server and client sides of the hosted application can be written in the same language (Python) and so that they may create and manage actors that exchange messages with one another, both locally and remotely, in a transparent manner and without the need to implement data conversions or other low-level mechanisms.

A presente pesquisa tem por objetivo desenvolver um sistema automático de detecção de eixos de caminhões a partir de imagens. Para isso, são apresentados dois sistemas automáticos: o primeiro para extração de imagens de caminhões a partir de filmagens de tráfego rodoviário feitas em seis locais de uma mesma rodovia situada no Estado de São Paulo, e o segundo, para detecção dos eixos dos caminhões nas imagens. Ambos os sistemas foram fundamentados em conceitos de Processamento de Imagens e Visão Computacional e o desenvolvimento foi feito utilizando programação em linguagem Python e as bibliotecas OpenCV e SciKit. O salvamento automático das imagens de caminhões foi necessário para a construção do banco de imagens utilizado no outro método: a detecção dos eixos dos veículos identificados. Neste estágio foram realizadas a segmentação da imagem do caminhão, a detecção propriamente dita e a classificação dos eixos. Na segmentação dos veículos, utilizou-se as técnicas de limiarização adaptativa seguida de morfologia matemática e em outra ocasião, o descritor de texturas LBP; enquanto na detecção, a Transformada de Hough. Da análise de desempenho desses métodos, a taxa de salvamento das imagens foi 69,2% considerando todos os caminhões que se enquadraram nos frames. Com relação às detecções, a segmentação das imagens dos caminhões feita utilizando limiarização adaptativa com morfologia matemática ofereceu resultados de 57,7% da detecção do total de eixos dos caminhões e 65,6% de falsas detecções. A técnica LBP forneceu, para os mesmos casos, respectivamente, 68,3% e 84,2%. O excesso de detecção foi um ponto negativo dos resultados e pode ser relacionado aos problemas do ambiente externo, geralmente intrínsecos às cenas de tráfego de veículos. Dois fatores que interferiram de maneira significativa foram a iluminação e a movimentação das folhas e galhos das árvores devido ao vento. Desconsiderando esse inconveniente, derivado dos fatores recém citados, as taxas de acerto dos dois tipos de segmentação aumentariam para 90,4% e 93,5%, respectivamente, e as falsas detecções mudariam para 66,5% e 54,7%. Desse modo, os dois sistemas propostos podem ser considerados promissores para o objetivo proposto. / This research aims to develop an automatic truck axle detection system using images. Two automatic systems are presented: one for the extraction of truck images from road videos recorded in a São Paulo state highway and one for the axle detection on images. Both systems are based on Image Processing and Computational Vision techniques, with using programming in Python and OpenCV and SciKit libraries. The truck image extraction system was necessary for the creation of image base, to be used on the axle detection system. Thereunto, image segmentation, axle detection and classification on images were made. In segmentation was used adaptive threshold technique, followed by mathematical morphology and on another time, LBP texture descriptors; for detection, was used Hough Transform. Performance analysis on these methods wielded 69.2% on image save rate, on trucks entirely framed on the image. About axle detection, the truck image segmentation using a combination of adaptive threshold and mathematical morphology wielded 57.7% on axle detection, whilst achieving 65.6% of false detection. Using LBP wielded, on the same images, 68.3% on axle detection and 84.2% of false detection. These excesses was a negative result and can be related to intrinsic issues on the external road traffic environment. Two main factors affected the result: lighting condition changes and the movement of tree leaves and branches. Disregarding these two factors, the proposed system had 90.4% axle truck detection rate using adaptive threshold and mathematical morphology and 93.5% using LBP, and the false detection, changed for 66.5% e 54.7%. Thus, both proposed systems are considered promising.

O Presente trabalho é o estudo das funcionalidades de um descritor e sua aplicabilidade em análise musical tendo como ponto de partida a obra a Ladainha de Nossa Senhora de Faustino Xavier do Prado. Utilizaremos os intervalos melódicos existentes na Ladainha como padrões para verificar similaridades nos contextos; Brasil, Itália e Portugal. Utilizaremos como ferramenta de busca de padrões o software MelodicMatch e desenvolveremos um software para tratar os resultados obtidos organizando-os em uma proposta metodológica que estará descrita na própria aplicação. / The present work is the study of the features of a descriptor and its use in musical analysis taking as its starting point the work of the Ladainha de Nossa Senhora - Faustino Xavier del Prado. We will use existing melodic intervals in the Ladainha as standards for checking similarities in the contexts; Brazil, Italy and Portugal. We will use as standards search tool MelodicMatch the software and develop software to handle the results organizing them into a methodology that will be described in the application itself.

No ciclo hidrológico, o processo de evapotranspiração (ET) representa a maior percentagem do total de precipitação. As maiores percentagens de ET são encontradas nas zonas ripárias visto a alta disponibilidade hídrica. Embora existam métodos acessíveis para estimativa da ET em zona ripária, como os baseados na flutuação diária do nível piezométrico, poucos estudos têm sido desenvolvidos devido à complexidade teórica da variável, ao elevado custo de equipamentos, e à reduzida largura dessas áreas. Desta forma, o objetivo deste trabalho foi avaliar 10 métodos baseados na flutuação diária do nível piezométrico usando o método de balanço de energia &#8211; Razão de Bowen. Para tal fim, foram construídos cinco poços e uma torre de monitoramento em uma área de zona ripária da bacia hidrográfica do Ribeirão da Onça (Brotas, São Paulo). Os dados dos poços e da torre de monitoramento foram coletados, a cada 15 minutos, no período de, respectivamente, abril de 2018 e junho de 2018, a janeiro de 2019. Para o processamento dos dados, foi feito um programa em linguagem Python. Nos dados do método de balanço de energia Razão de Bowen foram utilizados o critério de propagação de erros para a rejeição de estimativas, e o método de amostragem direta para a posterior reconstrução da série temporal. No período seco e chuvoso, as taxas médias de ET obtidas pelo método de balanço de energia &#8211; Razão de Bowen foram, respectivamente, de 3,62 &#177; 0,93 mm/dia e 5,09 &#177; 2,34 mm/dia. Entre os métodos baseados na flutuação diária do nível piezométrico, o método de Gribovski apresentou as melhores métricas na avaliação (<font style=\"text-decoration: overline;\"> &#961; = 0.69, <font style=\"text-decoration: overline;\"> &#964;s = 0.55, <font style=\"text-decoration: overline;\"> MAE = 1.77 mm/dia, <font style=\"text-decoration: overline;\"> RMSE = 2.07 mm/dia e <font style=\"text-decoration: overline;\"> MBE = 1.63 mm/dia), o qual mostra o potencial desse tipo de métodos para estimar a ET em zonas ripárias. / Evapotranspiration rate (ET) represents a significant percentage of precipitation in the water balance. At the riparian zones, ET slightly increases due to high water availability. Although there are available methods to ET estimation, such as the based on groundwater level variation, few studies have developed because of theorical complexity, high equipment cost and narrow band of the riparian zones. Thus, this work aims to evaluate 10 groundwater level variation methods using the energy balance &#8211; Bowen ratio method. For that, five wells and one monitoring tower were installed in a riparian zone of the Onça Creek basin (Brotas, São Paulo, Brazil). Data from wells and monitoring tower were collected at each 15 minutes from, respectively, April/2018 and June/2018 to January/2019. For data processing, a Python languague programme was developed. Error propagation method was applied to reject unreliable estimation of ET. Also, direct sampling was used to temporal serie rebuilding. The ET estimations from Bowen ration method range from 3,62 &#177; 0,93 mm/day and 5,09 &#177; 2,34 mm/day (dry and wet period, respectively). Among the groundwater level variation methods, the Gribovski method reached the smallest error (<font style=\"text-decoration: overline;\"> &#961; = 0.69, <font style=\"text-decoration: overline;\"> &#964; s = 0.55, <font style=\"text-decoration: overline;\"> MAE = 1.77 mm/day, <font style=\"text-decoration: overline;\"> RMSE = 2.07 mm/day e <font style=\"text-decoration: overline;\"> MBE = 1.63 mm/day). Then, the Griboski method may be used to ET estimation in riparian zones.