Property | Value |
Name | NCL-TR-2009008 |
Description |
Inductive Linkage Identification: Scalability, Robustness, and Population Sizing
Abstract: This paper proposes a linkage identification algorithm, named inductive linkage identification (ILI), to identify linkages, which are referred to as the interdependencies among variables. The proposed algorithm utilizes the ID3 decision tree to extract sets of variables based on the mutual relevance on the fitness differences caused by perturbation. In this study, we concentrate on the properties and characteristics of ILI, including its scalability, robustness, and population requirement. According to the experimental results, compared to other linkage learning techniques, ILI exhibits equal or better flexibility, scalability, and robustness. A theoretical population sizing model is also developed in this paper to reveal the population requirement for ILI to operate. The proposed population sizing model well agrees with the experimental results and such a model may provide an insight into perturbation-based as well as entropy-based linkage learning methodologies. |
Filename | NCL-TR-2009008.pdf |
Filesize | 350.53 kB |
Filetype | pdf (Mime Type: application/pdf) |
Creator | ypchen |
Created On: | 11/29/2009 20:18 |
Viewers | Everybody |
Maintained by | Editor |
Hits | 2384 Hits |
Last updated on | 12/09/2010 12:41 |
Homepage |