Communication Analysis and Privacy in CAI Based on Data Mining and Federated Learning

Abstract

Due to the fact that client data do not need to leave the local area, a distributed machine learning framework can aggregate training from several clients while preserving data privacy. In this essay, the development of CAI both domestically and internationally is reviewed and summarized, and the current state of CAI is examined. Communication analysis has so far been a key academic and theoretical area in federated learning, and some theoretical contributions have become the crucial theoretical foundations for understanding, defending, and guiding various human social behaviors. The major objective of knowledge distillation based on model responses is to provide students the ability to rapidly replicate the teacher’s model’s output. The experimental results demonstrate that the optimized Smith Regan model adopts the “Smith Logan” teaching design model, selects the courseware structure and record preservation as the teaching content in the fundamental CAI courseware design, and optimizes the teaching design from the perspectives of learning environment analysis, learner characteristics analysis, etc. Based on this, the model’s accuracy and robustness are increased by 7.34%.

Keywords

• federated learning
• CAI
• communication analysis
• optimization of Smith–Regan model