Stability and Performances in Biclustering Algorithms

Abstract. Stability is an important property of machine learning algorithms. Stability in clustering may be related to clustering quality or ensemble diversity, and therefore used in several ways to achieve a deeper understanding or better confidence in bioinformatic data analysis. In the specific field of fuzzy biclustering, stability can be analyzed by porting the definition of existing stability indexes to a fuzzy setting, and then adapting them to the biclustering problem. This paper presents work done in this direction, by selecting some representative stability indexes and experimentally verifying and comparing their properties. Experimental results are presented that indicate both a general agreement and some differences among the selected methods.

• Type: Conference
• Year: 2008
• Where: CIBB
• Authors: Maurizio Filippone, Francesco Masulli, Stefano Rovetta
• Keywords: Artificial Intelligence, CIBB 2008, Machine Learning Algorithms, Representative Stability Indexes, Stability Indexes