Controlling Self-Organization and Handling Missing Values in SOM and GTM

Abstract

In this paper, we study fundamental properties of the Self-Organizing Map (SOM) and the Generative Topographic Mapping (GTM), ramifications of the initialization of the algorithms and properties of the algorithms in presence of missing data. We show that the commonly used principal component analysis (PCA) initialization of the GTM does not guarantee good learning results with complex, high-dimensional data. We propose initializing the GTM with SOM and demonstrate usefulness of this improvement using the ISOLET data set. We also propose a revision to the batch SOM algorithm called the Imputation SOM and show that the new algorithm is more robust in presence of missing data. We compare the performance of the algorithms in the missing value imputation task. We also announce a revised version of the SOM Toolbox for Matlab with added GTM functionality.KeywordsData VectorReference VectorNeighborhood FunctionCombine ErrorGood Learning ResultThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.