5G+ Smart Classroom: Digital Innovation and Optimization of Orthopedic Practical Teaching

Guichun Zhang, Xinru Li, Guilai Zuo, Chenglin Sang

Abstract: This study utilizes CiteSpace, a bibliometric tool, to map the evolution of Karl Marx's influence on philosophy from 2015 to the present, focusing on the intellectual shifts and emerging trends in Marxist thought. By analyzing co-citation networks and keyword bursts, the research uncovers the growing prominence of themes such as eco-Marxism, Marx party, and spirituality. The results reveal an increasing intersection between Marxist philosophy and contemporary challenges like environmental sustainability, political movements, and cultural dimensions of capitalist critique. The study highlights a shift toward interdisciplinary engagement, with Marxist ideas being applied to both traditional and novel fields, extending the relevance of Marx’s philosophy to modern-day issues. Future research is recommended to continue exploring these emerging topics, particularly the intersection of Marxism with ecological crises and the digital economy.

: Download (free)  CSE-PDF

Prediction of impact Sensitivity of Aromatic Compounds with Nitro Group Based on QSPR

Zixian Yang, Peijian Zhang

Abstract:

Nitroaliphatic compounds constitute an important class of energetic materials, and their impact sensitivity (h_50%) is a key parameter for evaluating the safety of these compounds under mechanical shock. In this study, nonlinear quantitative structure-property relationship (QSPR) models are developed to accurately predict the impact sensitivity of nitroaliphatic compounds, providing a reliable alternative to experimental testing for safety assessment of energetic materials.

Three key molecular descriptors previously screened by Prana are employed to construct four nonlinear QSPR models: random forest (RF), linear mixed-kernel support vector regression (LMIX-SVR), Gaussian process regression (GPR), and linear multilayer perceptron (L-MLP). Model performances are comprehensively evaluated through internal validation, external validation, and applicability domain analysis.

Overall, models based on kernel methods with smooth distance metrics (GPR and LMIX-SVR) or lightweight regularized neural networks (L-MLP) demonstrate superior fitting ability and prediction robustness. Among them, the L-MLP model exhibits the best overall performance, with a coefficient of determination of R_Train² = 0.9040 and root-mean-square error RMSE_Train = 0.1406 for the training set; for the external validation set, R_EXT² = 0.8343 and RMSE_EXT = 0.1720; and cross-validation results of Q_LOO² = 0.8511 and Q_5CV² = 0.8526.

Within the applicability domain, for the internal validation set, the L-MLP model achieves R_IN² = 0.8073 and RMSE_IN = 0.1776, demonstrating significantly higher prediction accuracy than the conventional multiple linear regression model. These results indicate that nonlinear QSPR models based on simple structural descriptors can effectively capture the complex structure–property relationships governing impact sensitivity.

In particular, the L-MLP model combines the interpretability of linear models with the nonlinear fitting capability of neural networks, offering a reliable methodological reference for rapid prediction of the impact sensitivity of nitroaliphatic compounds and for safety screening of energetic materials.

: Download (free)  CSE-PDF