Abstract
As the major cause of deaths worldwide, cardiovascular diseases are responsible for about 17.9 million deaths per year 1. Research on new technologies and methodologies allowed the acquisition of reliable data in several high income countries, however, in various developing countries, due to poverty and common scarcity of resources, this has not been reached yet. In this work, cardiovascular data acquired using cardiac auscultation is going to be used to detect cardiac murmurs through an innovative deep learning approach. The proposed screening algorithm was built using pre-trained models comprising Residual Neural Networks, namely Resnet50, and Visual Geometry Groups, such as VGG16 and VGG19. Furthermore, and up to our knowledge, our proposal is the first one that characterizes heart murmurs based on their frequency components, i.e. the murmur pitch. Such analysis may be used to augment the system's capability on detecting heart diseases. A novel decision-making function was also proposed regarding the murmur's pitch. From our experiments, low-pitch murmurs were more difficult to detect, with final f1-score values nearing the 0.40 value mark for all three models, while high-pitch murmurs presented an higher f1-score value of about 0.80. This might be due to the fact that the low-pitch share their respective frequency range with the normal and fundamental heart sounds, therefore making it harder for the model to correctly detect their presence whereas high-pitch murmurs' frequencies distance from the latter.

Abstract
Regression testing is a software engineering maintenance activity that involves re-executing test cases on a modified software system to check whether code changes introduce new faults. However, it can be time-consuming and resource-intensive, especially for large systems. Regression testing selection techniques can help address this issue by selecting a subset of test cases to run. The change-based technique selects a subset of test cases based on the modified software classes, reducing the test suite size. Thereby, it will cover a smaller number of classes, decreasing the efficiency of the test suite to reveal design flaws. From this perspective, code smells are known to identify poor design and threaten the quality of software systems. In this study, we propose an approach to combine code change and smell to select regression tests and present two new techniques: code smell based and code change and smell. Additionally, we developed the Regression Testing Selection Tool (RTST) to automate the selection process. We empirically evaluated the approach in Defects4J projects by comparing the new techniques' effectiveness with the change-based as a baseline. The results show that the change-based technique achieves the highest reduction rate in the test suite size but with less class coverage. On the other hand, test cases selected using code smells and changed classes combined can potentially find more bugs. The code smell-based technique provides a comparable class coverage to the code change and smell approach. Our findings highlight the benefits of incorporating code smells in regression testing selection and suggest opportunities for improving the efficiency and effectiveness of regression testing.

Abstract
A common problem in distribution planning is the scarcity of historic data (training examples) relative to the number of variables, meaning that most data-driven techniques cannot be applied in such situations, due to the risk of overfitting. Thus, the suitable regression techniques are restrained to efficient models, preferably with embedded regularization features. This article compares three of these techniques: LASSO, Bayesian and CMLR (Conditioned multi-linear regression - a new approach developed within the scope of a project with a distribution company). The results showed that each technique has its own advantages and limitations. The Bayesian regression has the main advantage of providing inherent confidence intervals. The LASSO is a very economic and efficient regression tool. The CMLR is versatile and provided the best performance.A common problem in distribution planning is the scarcity of historic data (training examples) relative to the number of variables, meaning that most data-driven techniques cannot be applied in such situations, due to the risk of overfitting. Thus, the suitable regression techniques are restrained to efficient models, preferably with embedded regularization features. This article compares three of these techniques: LASSO, Bayesian and CMLR (Conditioned multi-linear regression - a new approach developed within the scope of a project with a distribution company). The results showed that each technique has its own advantages and limitations. The Bayesian regression has the main advantage of providing inherent confidence intervals. The LASSO is a very economic and efficient regression tool. The CMLR is versatile and provided the best performance.

Abstract
The adoption of third-party machine learning (ML) cloud services is highly dependent on the security guarantees and the performance penalty they incur on workloads for model training and inference. This paper explores security/performance trade-offs for the distributed Apache Spark framework and its ML library. Concretely, we build upon a key insight: in specific deployment settings, one can reveal carefully chosen non-sensitive operations (e.g. statistical calculations). This allows us to considerably improve the performance of privacy-preserving solutions without exposing the protocol to pervasive ML attacks. In more detail, we propose Soteria, a system for distributed privacy-preserving ML that leverages Trusted Execution Environments (e.g. Intel SGX) to run computations over sensitive information in isolated containers (enclaves). Unlike previous work, where all ML-related computation is performed at trusted enclaves, we introduce a hybrid scheme, combining computation done inside and outside these enclaves. The experimental evaluation validates that our approach reduces the runtime of ML algorithms by up to 41% when compared to previous related work. Our protocol is accompanied by a security proof and a discussion regarding resilience against a wide spectrum of ML attacks.

Abstract
The shotcrete process has been extensively used for many years in different civil and mining operations. Nevertheless, it is still either applied by an operator which controls the shotcrete nozzle manually or through a remote control. In either case, the operation is entirely controlled by the operator. Automating the shotcrete process involves developments in different parts of the process, such as the tunnel scanning for 3D model generation and the shotcrete path automatic generation and execution. This paper describes the work developed for this last part, namely the automatic generation and execution of a shotcrete path, given the mesh of a tunnel and a set of input parameters, for application in railway tunnels. The developed path also considers specificities of the concrete projection process, such as the uncontrolled flow variation due to the pumping systems, generating a trajectory that aims at minimizing this effect. Results are shown using a realistic simulator and an uneven railway tunnel, using an industrial robot mounted on a railway wagon.

Abstract
Educational chatbots are digital tools designed to assist learners in various educational settings. These chatbots use natural language processing (NLP) and machine learning algorithms to simulate human conversation and respond to user queries in a way that facilitates learning. They can be integrated into various educational platforms such as learning management systems, educational apps, and websites to provide learners with a personalized and interactive learning experience. Our paper discusses different scenarios for educational purposes and suggests in total four scenarios for educational needs.