Fuzzy relevance vector machine for learning from unbalanced data and noise

Handing unbalanced data and noise are two important issues in the field of machine learning. This paper proposed a complete framework of fuzzy relevance vector machine by weighting the punishment terms of error in Bayesian inference process of relevance vector machine (RVM). Above problems can be learned within this framework with different kinds of fuzzy membership functions. Experiments on both synthetic data and real world data demonstrate that fuzzy relevance vector machine (FRVM) is effective in dealing with unbalanced data and reducing the effects of noises or outliers.

• Type: Journal
• Year: 2008
• Where: PRL
• Authors: Dingfang Li, Wenchao Hu, Wei Xiong, Jin-Bo Yang
• Keywords: Bayesian Inference Process, Fuzzy Relevance Vector, PRL 2008, Relevance Vector Machine