Using Artificial Intelligence (AI) and machine learning technologies to automatically mine latent patterns from educational data holds great potential to inform teaching and learning practices. However, the current AI technology mostly works as "black box"-only the inputs and the corresponding outputs are available, which largely impedes researchers from gaining access to explainable feedback. This interdisciplinary work presents an explainable AI prototype with visualized explanations as feedback for computer-supported collaborative learning (CSCL). This research study seeks to provide interpretable insights with machine learning technologies for multimodal learning analytics (MMLA) by introducing two different explanatory machine learning-based models (neural network and Bayesian network) in different manners (end-to-end learning and probabilistic analysis) and for the same goal-provide explainable and actionable feedback. The prototype is applied to the real-world collaborative learning scenario with data-driven learning based on sensor-data from multiple modalities which can assess collaborative learning processes and render explanatory real-time feedback.@en

A hybrid lattice Boltzmann method–very-large-eddy simulation (LBM-VLES) solver for high-speed nonisothermal subsonic flows is used to simulate the unsteady jet flow exhausting from a single axi-symmetric nozzle, as well as the associated noise spectra and directivity. The jet exit Mach number and temperature ratio are set according to three representative operating conditions from the NASA SMC000 experimental campaign. The farfield noise is computed through a Ffowcs Williams and Hawkings analogy applied to a fluid surface encompassing the jet plume. Both time- and frequency-domain formulations are used, the latter in combination with an azimuthal Fourier transform of the linear source terms to analyze the contribution of the different azimuthal components. A resolution study is carried out for both aerodynamic and acoustic results. The near- and far-field results confirm that the underlying flow features and noise mechanisms are fully represented by the numerical solution. A wavelet decomposition technique is applied to analyze the source mechanisms for a heated core case. This is achieved by separating the coherent flow motion from the chaotic perturbations in the turbulent flows. Finally, a frequencydomain integral formulation is used to analyze the acoustic far-field of the two segregated components. @en

To address the problem of training on small datasets for action recognition tasks, most prior works are either based on a large number of training samples or require pre-trained models transferred from other large datasets to tackle overfitting problems. However, it limits the research within organizations that have strong computational abilities. In this work, we try to propose a data-efficient framework that can train the model from scratch on small datasets while achieving promising results. Specifically, by introducing a 3D central difference convolution operation, we proposed a novel C3D neural network-based two-stream (Rank Pooling RGB and Optical Flow) framework for the task. The method is validated on the action recognition track of the ECCV 2020 VIPriors challenges and got the 2nd place (88.31%). It is proved that our method can achieve a promising result even without a pre-trained model on large scale datasets. The code will be released soon. @en