Skip to Main Content
This paper has proposed a new variance-based sorting initial pattern library algorithm for machine learning. First, we sort the training vector set based on vector variance; second, categorize it to several subsets with variance thresholds; last, select some number of pattern vectors from the subsets to form the initial pattern library. This new initial pattern library algorithm is tested by two unsupervised machine learning algorithms: self-organizing feature maps (SOM) algorithm and frequency sensitive self-organizing feature maps (FSOM) algorithm. Experimental results for image coding show that this new initial pattern library algorithm is better than the common random sampling algorithm.