Particle Swarm Optimization for Automatic Selection of Relevance Feedback Heuristics

摘要

Relevance feedback (RF) is an iterative process which refines the retrievals by utilizing user's feedback marked on retrieved results. Recent research has focused on the optimization for RF heuristic selection. In this paper, we propose an automatic RF heuristic selection framework which automatically chooses the best RF heuristic for the given query. The proposed method performs two learning tasks: query optimization and heuristic-selection optimization. The particle swarm optimization (PSO) paradigm is applied to assist the learning tasks. Experimental results tested on a content-based retrieval system with a real-world image database reveal that the proposed method outperforms several existing RF approaches using different techniques. The convergence behavior of the proposed method is empirically analyzed.