作者: D. Wienke , D. Domine , L Buydens , J. Devillers
DOI: 10.1016/B978-012213815-7/50007-8
关键词:
摘要: The family of adaptive resonance theory (ART) based artificial neural networks is concerned with distinct theoretical models for unsupervised and supervised pattern recognition. aim the paper presented in this chapter to explore interest ART algorithms quantitative structure–activity relationships (QSAR). In after a brief presentation neurophysiological basis different paradigms, ART-2a FuzzyART are presented, heuristic potency these methods illustrated from two case studies. former dealing headspace analysis roses clearly shows didactic manner how set parameters results can be interpreted. second study real-life application showing techniques optimal test series selection. compared those obtained hierarchical cluster visual mapping methods. advantages drawbacks each method discussed. Furthermore, features methodology such as rapid training speed, self-organization behavior, interpretability network weights also