CLASSIFICATION AND CLUSTERING - USING NEURAL NETWORKS

The connectionist approach uses two learning paradigms: supervised and unsupervised. Neutral networks which are based on these two different types of learning paradigms have been tested for their classification and clustering abilities in normal conditions as well as in hostile c onditions. We combined both paradigms together with feature enhancemen t, and the test case has been chosen from the field of nuclear magneti c resonance (NMR) spectroscopy. Our research indicates that neural net works based upon both paradigms together with feature enhancements exh ibit better capabilities than the neutral networks based on a single p aradigm alone without feature enhancement. We also observed that neura l network tuning is one of the most difficult aspects of the connectio nist pos approach. For achieving optimum results, some of the suggesti ons are incorporated. The results obtained from classification and clu stering are also shown in this paper.