Abstract
<jats:p>Addressing the multidimensionality of skills and nonlinearity of learning rhythm in the guzheng learning process, this paper proposes a personalized teaching path adaptive planning algorithm (DRL-PGT) combining deep reinforcement learning. This algorithm models the learner’s performance behavior as a state space, uses deep neural networks for feature extraction and policy optimization, and dynamically adjusts the difficulty of the teaching task based on a reward function. Experimental results show that the DRL-PGT algorithm improves the average reward value by 42.2% and 18.2% compared to traditional Q-learning and DQN methods, respectively, and reduces the number of convergence rounds by 55.6%, demonstrating superior learning efficiency and stability. In actual teaching experiments, the performance accuracy increased from 82.4% to 93.8%, rhythmic stability improved by 36.4%, and student satisfaction reached 4.7 points. The research results verify the effectiveness of deep reinforcement learning in optimizing personalized teaching paths for guzheng and can provide a scalable technical path for the intelligent teaching of traditional Chinese musical instruments.</jats:p>