A Student Data Generation Model Based on WGAN and Multi-Head Attention Mechanism
The evaluation of students' interdisciplinary ability and innovation ability is the focus and difficulty of current research. In view of the scarcity of public student data and the difficulty to obtain, this paper proposes a student data generation model based on Wasserstein Generative Adversarial Networks ( WGAN ) and multi-head attention mechanism. Different from traditional Generative Adversarial Networks ( GAN) , WGAN takes Wasserstein distance as the objective function, strengthens the stability of the generator and the convergence of training, and improves the quality of the generated data. To address the possible problems of slow convergence and generation of low-quality data in WGAN, a multi-attention mechanism is introduced into the generator of WGAN to better capture the potential patterns and structures in student data, and improve the quality of the generated data. Furthermore, mean value, standard deviation, and median are used to objectively evaluate the generated data. The experiment results show that the presented method can generate higher quality student data.
万方数据
维普
