A multi-instance ensemble learning model based on concept lattice

✍ Scribed by Xiangping Kang; Deyu Li; Suge Wang

• Publisher: Elsevier Science
• Year: 2011
• Tongue: English
• Weight: 311 KB
• Volume: 24
• Category: Article
• ISSN: 0950-7051
• DOI: 10.1016/j.knosys.2011.05.010

✦ Synopsis

This paper introduces concept lattice and ensemble learning technique into multi-instance learning, and proposes the multi-instance ensemble learning model based on concept lattice which can be applied to content-based image retrieval, etc. In this model, a }-concept lattice is built based on training set firstly. Because bags rather than instances in bags will serve as objects of formal context in the process of building }-concept lattice, the corresponding time complexity and space complexity can be effectively descend to a certain extent; Secondly, the multi-instance learning problem is divided into multiple local multi-instance learning problems based on }-concept lattice, and local target feature sets are found further in each local multi-instance learning problem. Finally, the whole training set can be classified almost correctly by ensemble of multiple local target feature sets. Through precise theorization and extensive experimentation, it proves that the method is effective. Conclusions of this paper not only help to understand multi-instance learning better from the prospective of concept lattice, but also provide a new theoretical basis for data analysis and processing.