Suchergebnisse

Deep learning evaluation is a new direction formed by the intersection of multiple domains, and the core issue is how to visualize collaborative learning models to motivate learners. Therefore, this paper realizes real-time knowledge sharing and facilitates learners' interaction through computer-supported collaborative learning (CSCL) technology. In this paper, we collect, label, and analyze data based on five modalities: brain, behavior, cognition, environment, and technology. In this paper, a computer-supported collaborative learning process analysis model is developed under the threshold of multimodal data analysis. The model is based on roles and CSCL for intelligent network collaboration. This paper designs and develops an interactive visualization tool to support online collaborative learning process analysis. In addition, this paper conducts a practical study in an online classroom. The results show that the model and the tool can be effectively used for online collaborative learning process analysis, and the test model results fit well. The entropy index of the test model took a value of about 0.85, and about less than 10% of the individuals were assigned to the wrong profile. During the test, the participation of participants gradually increased from 5% to about 25%, and the participation effect improved by about 80%. This indicates the strong applicability value of the computer-supported collaborative learning process analysis model under the multimodal data analysis perspective.

Recently, issues of climate change, environment abnormality, individual requirements, and national defense have caused extensive attention to the commercial, scientific, and military development of unmanned surface vehicles (USVs). In order to design high-quality routes for a multi-sensor integrated USV, this work improves the conventional particle swarm optimization algorithm by introducing the greedy mechanism and the 2-opt operation, based on a combination strategy. First, a greedy black box is established for particle initialization, overcoming the randomness of the conventional method and excluding a great number of infeasible solutions. Then the greedy selection strategy and 2-opt operation are adopted together for local searches, to maintain population diversity and eliminate path crossovers. In addition, Monte-Carlo simulations of eight instances are conducted to compare the improved algorithm with other existing algorithms. The computation results indicate that the improved algorithm has the superior performance, with the shortest route and satisfactory robustness, although a fraction of computing efficiency becomes sacrificed. Moreover, the effectiveness and reliability of the improved method is also verified by its multi-sensor-based application to a USV model in real marine environments.