Learned from Neural Networks
Robert P. W. Duin · 2000
Many problems in data analysis and pattern recognition may be attacked by neural networks. Sometimes this approach is better, sometimes it is worse than the use of alternatives. A general discussion is presented on possibilities, advantages and disadvantages of their use in comparison with more specific approaches. The study of neural networks, once almost a `black box' has given a better understanding of the possibilities of pattern learning. The recent development of more specific methods has been strongly stimulated by this knowledge.