Abstract
Diabetes has significant effects on the human body, one of which is the increase in the blood pressure and when not diagnosed early, can cause severe vision complications and even lead to blindness. Early screening is the key to overcoming such issues which can have a significant impact on rural areas and overcrowded regions. Mobile systems can help bring the technology to those in need. Transfer learning based Deep Learning algorithms combined with mobile retinal imaging systems can significantly reduce the screening time and lower the burden on healthcare workers. In this paper, several efficiency factors of Diabetic Retinopathy detection systems based on Convolutional Neural Networks are tested and evaluated for mobile applications. Two main techniques are used to measure the efficiency of DL based DR detection systems. The first method evaluates the effect of dataset change, where the base architecture of the DL model remains the same. The second method measures the effect of base architecture variation, where the dataset remains unchanged. The results suggest that the inclusivity of the datasets, and the dataset size significantly impact the DR detection accuracy and sensitivity. Amongst the five chosen lightweight architectures, EfficientNet-based DR detection algorithms outperformed the other transfer learning models along with APTOS Blindness Detection dataset. © 2023, ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering.

Abstract
STEM areas (Science, Technology, Engineering and Math) are continuously growing but the number of technical workers do not accompany that growth. As the 21st century brings new challenges, students should be prepared for an increasingly complex life and work environments that will privilege proficiency in Learning and Innovation Skills that include Creativity and Innovation, Critical Thinking and Problem Solving, Communication and Collaboration. Also, the need to continuously explore new pedagogical practices in teaching and learning creates an opportunity to build new contents by balancing a stable and tested curriculum with new tools that stimulate creativity, allowing students to better understand the world they live in. This article describes the development of an educational robotics kit, aimed at children and teens from 8 to 18 years old, meant to work as an interdisciplinary teaching tool that can be applied directly in a curriculum.

Abstract
Secure multiparty computation (SMC) allows for complex computations over encrypted data. Privacy concerns for cloud applications makes this a highly desired technology and recent performance improvements show that it is practical. To make SMC accessible to non-experts and empower its use in varied applications, many domain-specific compilers are being proposed.We review the role of these compilers and provide a formal treatment of the core steps that they perform to bridge the abstraction gap between high-level ideal specifications and efficient SMC protocols. Our abstract framework bridges this secure compilation problem across two dimensions: 1) language-based source- to target-level semantic and efficiency gaps, and 2) cryptographic ideal- to real-world security gaps. We link the former to the setting of certified compilation, paving the way to leverage long-run efforts such as CompCert in future SMC compilers. Security is framed in the standard cryptographic sense. Our results are supported by a machine-checked formalisation carried out in EasyCrypt.

Abstract
Ordinal classification tasks are present in a large number of different domains. However, common losses for deep neural networks, such as cross-entropy, do not properly weight the relative ordering between classes. For that reason, many losses have been proposed in the literature, which model the output probabilities as following a unimodal distribution. This manuscript reviews many of these losses on three different datasets and suggests a potential improvement that focuses the unimodal constraint on the neighborhood around the true class, allowing for a more flexible distribution, aptly called quasi-unimodal loss. For this purpose, two constraints are proposed: A first constraint concerns the relative order of the top-three probabilities, and a second constraint ensures that the remaining output probabilities are not higher than the top three. Therefore, gradient descent focuses on improving the decision boundary around the true class in detriment to the more distant classes. The proposed loss is found to be competitive in several cases.

Abstract
In recent years, machine learning techniques, especially genetic programming (GP), have been a powerful approach for automated design of the priority rule-heuristics for the resource-constrained project scheduling problem (RCPSP). However, it requires intensive computing effort, carefully selected training data and appropriate assessment criteria. This research proposes a GP hyper-heuristic method with a duplicate removal technique to create new priority rules that outperform the traditional rules. The experiments have verified the efficiency of the proposed algorithm as compared to the standard GP approach. Furthermore, the impact of the training data selection and fitness evaluation have also been investigated. The results show that a compact training set can provide good output and existing evaluation methods are all usable for evolving efficient priority rules. The priority rules designed by the proposed approach are tested on extensive existing datasets and newly generated large projects with more than 1,000 activities. In order to achieve better performance on small-sized projects, we also develop a method to combine rules as efficient ensembles. Computational comparisons between GP-designed rules and traditional priority rules indicate the superiority and generalization capability of the proposed GP algorithm in solving the RCPSP.

Abstract
Robotics competitions encourage the development of solutions to new challenges that emerge in sync with the rise of Industry 4.0. In this context, robotic simulators are employed to facilitate the development of these solutions by disseminating knowledge in robotics, Education 4.0, and STEM. The RobotAtFactory 4.0 competition arises to promote improvements in industrial challenges related to autonomous robots. The official organization provides the simulation scene of the competition through the open-source SimTwo simulator. This paper aims to integrate the SiwTwo simulator with the Robot Operating System (ROS) middleware by developing a framework. This integration facilitates the design of robotic systems since ROS has a vast repository of packages that address common problems in robotics. Thus, competitors can use this framework to develop their solutions through ROS, allowing the simulated and real systems to be integrated.