Effectiveness of Value Granulation in Machine Learning for Massively Large and Complex Domain

Zachary I Moore, Atsushi Inoue.

Considered from data analysis and dynamic optimization view point, computer networks are massively large and complex data domains where analytical computation (e.g. statistics) is, despite its needs, often found infeasible. In an attempt to address the issues raised by such domains, we are currently studying Granular Computing, a newly emerging paradigm, and its application to Machine Learning. This paper reports effectiveness of value granulation (such as discretization and quantization) in Machine Learning from aspects of complexity reduction, learning capability, and intelligent system development.

PDF full paper