Limitations of Competitive Learning, neural-network,pdfs, lecture-notes, downloads
Competitive learning is lacking in the capability to add modern clusters whenever deemed necessary.
Competitive learning does not guarantee stability in forming clusters. If the learning rate η is constant, so the winning unit that responds to a pattern may well continue altering during training.
If the learning rate η is minimizing with time, it might become too small to update cluster centers when new data of differentprobability are presented.