Initialization of Self-Organizing Maps: Principal Components Versus Random Initialization. A Case Study
This work addresses initialization methods for SOMs, which is an incremental improvement for practitioners in data analysis and machine learning.
The study compared random initialization (RI) and principal component initialization (PCI) for Self-Organizing Maps, finding that RI performed better for non-linear datasets, while PCI's performance on quasi-linear datasets was inconclusive.
The performance of the Self-Organizing Map (SOM) algorithm is dependent on the initial weights of the map. The different initialization methods can broadly be classified into random and data analysis based initialization approach. In this paper, the performance of random initialization (RI) approach is compared to that of principal component initialization (PCI) in which the initial map weights are chosen from the space of the principal component. Performance is evaluated by the fraction of variance unexplained (FVU). Datasets were classified into quasi-linear and non-linear and it was observed that RI performed better for non-linear datasets; however the performance of PCI approach remains inconclusive for quasi-linear datasets.