Practical Teaching Reform of the “Deep Learning and Applications” Course Through the Collaboration of Knowledge Graphs and AI Agents

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Addressing issues such as high programming barriers, fragmented resources, delayed feedback, and limited pathways for innovation in the practical instruction of the “Deep Learning and Applications” course, this paper proposes a reform plan for practical teaching centered on a knowledge graph and featuring generative AI agents as bidirectional interfaces. This proposal leverages the ChaoXing platform to construct a course knowledge graph comprising three types of nodes—conceptual, operational, and problem-based—and utilizes generative AI to develop a teaching agent for code practice. This forms a closed-loop support system encompassing four stages: semantic parsing, code framework generation, error diagnosis, and case expansion. Consequently, it establishes two implementation pathways: precise student assistance and refined teaching support for instructors, complemented by multi-stakeholder collaborative evaluation and AI usage constraints. This plan has undergone two rounds of iterative implementation, accumulating process-based data that can serve as a reference for practical teaching reforms in artificial intelligence courses.