The disconnect between static knowledge graphs and the individualized, dynamic learning needs of the students renders the adaptive learning path generation less effective in AI-driven adaptive learning. This paper proposes a unique integrated model that combines dynamic graph attention networks and curriculum reinforcement learning to address this problem. The dynamic graph attention network builds a personalized representation of the knowledge graphs by creating contextualized graph representations of the knowledge graph nodes using the current status of the learner to create an individual learning experience for each student. The second element of the architecture is a curriculum reinforcement learning agent that uses proximal policy optimization as its method of action. The curriculum reinforcement agent takes the fused graph representation as its input state and the recommended knowledge points as its action space. This framework performs multi-objective policy optimization for the learning path using curriculum rewards that integrate immediate learning gains, long-term knowledge structure coverage, and cognitive load smoothness, thereby generating a learner’s dynamically generated adaptive pathway. Experiments show that the proposed method achieves an 85.6% mastery rate at the end of the generated path, improves learning efficiency by 23.7%, and achieves a knowledge graph coverage rate of 79.1%, effectively validating its comprehensive performance in improving learning outcomes, optimizing the learning process, and deepening the utilization of knowledge structures.
Yi et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: