Classification of geophysical features with CALM neural networks

Unsupervised CALM neural networks are used to classify geophysical features in quasi-real-time. Geophysical features are extracted from geophysical datasets using classical pre-processing techniques. Corresponding feature vectors, describing the properties of such geophysical features, are interactively reduced to vectors containing the most suitable features for the classification purpose. The importance of feature scaling is demonstrated by classifying unsealed and scaled feature vector sets. Further, it is shown that unsupervised learning of a small feature vector set is sufficient to quickly classify a novel dataset. Finally, unsupervised higher order geophysical phenomena are classified by hierarchical network, utilising contextual information between geophysical features