Topological Kernel Bayesian ARTMAP
Naoki Masuyama, Chu Kiong Loo, Hisao Ishibuchi, Yusuke Nojima, Yiping Lin · 2018
This paper proposes a robust supervised learning model based on ARTMAP framework by incorporating a topological clustering algorithm. The noise tolerance is one of the essential abilities for supervised learning because the supervised data from the practical environment may contain a certain level of noise information. The simulation experiment focuses on the supervised learning with incorrect supervised data for classification. Specifically, incorrect supervised data, which is defined by intentionally changing label data with an arbitrary ratio, is utilized in the learning sequence. The results show that the proposed model achieves superior noise tolerance ability comparing with the typical and state-of-the-art models of ARTMAP.