Effective discriminative TCM-KNN for incremental learning

摘要: Incremental learning is an efficient scheme for reducing computational complexity of batch learning. Label information in each update helpful to discriminative model incremental However, the procedure labeling samples always a time-consuming and tedious task. In this paper, we propose two algorithms unknown samples, one Transductive Confidence Machine K-Nearest Neighbor (TCM-KNN), other its improved algorithm choosing good quality enhancing performance samples; then these methods applied learning[2] before updating model. Experiment on PIE database has been carried out comparing their recognition rate complexity. Extensive experimental results show that proposed method more robust effective than