Abstract
Rival penalized competitive learning (RPCL) algorithm can automatically allocate an appropriate number of units for an input data set. However, RPCL algorithm randomly picks the initial cluster centers, which would significantly deteriorate its performance when the seeds are inappropriately selected. We propose an improved method in which the result of k-means is used to optimize the initial cluster centers. Moreover, RPCL algorithm randomly selects sample from data set, not considering the distribution of samples. The idea of regional density of samples is introduced to select samples according to the distribution of samples. Using algae images as real data and the box-counting dimension of them as the feature vectors set, we demonstrate the improved RPCL algorithm outperforms the conventional one.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call