We present 3DRegNet, a novel deep learning architecture for the registration of 3D scans. Given a set of 3D point correspondences, we build a deep neural network to address the following two challenges: (i) classification of the point correspondences into inliers/outliers, and (ii) regression of the motion parameters that align the scans into a common reference frame. With regard to regression, we present two alternative approaches: (i) a Deep Neural Network (DNN) registration and (ii) a Procrustes approach using SVD to estimate the transformation. Our correspondence-based approach achieves a higher speedup compared to competing baselines. We further propose the use of a refinement network, which consists of a smaller 3DRegNet as a refinement to improve the accuracy of the registration. Extensive experiments on two challenging datasets demonstrate that we outperform other methods and achieve state-of-the-art results. The code is available.

We construct thin-shell electrically charged wormholes in d-dimensional general relativity with a cosmological constant. The wormholes constructed can have different throat geometries, namely, spherical, planar, and hyperbolic. Unlike the spherical geometry, the planar and hyperbolic geometries allow for different topologies and in addition can be interpreted as higher-dimensional domain walls or branes connecting two universes. In the construction we use the cut-and-paste procedure by joining together two identical vacuum spacetime solutions. Properties such as the null energy condition and geodesics are studied. A linear stability analysis around the static solutions is carried out. A general result for stability is obtained from which previous results are recovered.

Higher education institutions are increasingly concerned with the professional insertion of graduates in the labour market and with the design of institutional mechanisms to facilitate students’ transition from higher education to work, particularly given the context of scarcity of financial resources and the rise of graduate unemployment. This issue has been addressed, inter alia, through the creation of study programmes with internships. Despite the public discourse encouraging the use of such strategies, there is a general consensus regarding the absence of empirical studies on the professional value of these strategies. This article aims to assess two interrelated questions: the extent to which measures of graduate unemployment rate tend to decrease after the introduction of internships in Portuguese study programmes; and the extent to which this effect applies to the different institutions that comprise the Portuguese tertiary education landscape. It also seeks to contribute to the debate on the relevance of the structure and nature of internships, which are factors frequently neglected in the literature.

In this study we sought to verify whether network analyses could be used to identify key players in attacking phases of a professional football match and establish the main interactions and preferential linkages between attacking teammates during competitive performance. For this purpose, we examined circulation of the ball on field during randomly selected attacking phases of play in a video-taped Portuguese Premier League match. We observed six matches and notated 1488 collective attacking actions, including: passes completed, passes received, and crosses involving a total of 4126 intra-team interactions (eg., 2063 passes and crosses performed and 2063 passes and crosses received). We used Amisco® software to perform quantitative and qualitative analyses of the attacking actions. Results indicated how key individual players are instrumental in orchestrating team performance, exerting a powerful influence in creating attacking patterns of play. Our findings may help coaches and sports scientists quantify the contributions and interactions of individual team members through analysis of their relevant actions in team sports like football.