Formalizing Adaptive Sampling for DBMS Performance Modeling: A Feedback-Driven Transfer Learning Pipeline

This paper presents a formalized approach to optimizing Database Management System performance prediction through an adaptive sampling pipeline. It integrates a feedback mechanism that dynamically guides data collection based on the learning progress within a transfer learning framework. By formalizing the separation of parameters and a two-stage sampling strategy, the system efficiently reduces data requirements while maintaining model accuracy. This paper AI ML (Modeling Feedback mechanism, formalizing the learning process) demonstrates how intelligent, feedback-driven data acquisition can significantly enhance the efficiency of performance modeling in complex database environments.