Learning dynamic user model in Bayesian image retrieval

Long-term learning exploiting historical RF information benefits CBIRS in both effectiveness and efficiency. However, distinguishing specific user understanding of image is a key issue deserving much attention. Based on the Bayesian framework in PicHunter, we propose a probabilistic model incorporating long-term learning to estimate a dynamic user model. By using RF sequence as the user pattern, our approach can gradually update the prediction of current user based on matching the current user pattern with the user patterns in Log according to Edit Distance. Compared with the invariant user model in PicHunter, our model is capable of dynamically adjusting when more user actions are observed, thus provide more accurate prediction for probability distribution. Experimental results show that our approach can improve the retrieval effectiveness apparently.