抄録
This paper deals with the problem of information compression, similarity analysis and classification of large information sets that may represent various operating conditions of machines or different digital images. Here two unsupervised learning algorithms for information compression are presented, namely the standard off-line learning with fixed number of neurons and the on-line learning algorithm. The latter one is much faster and therefore convenient for information compression of large data sets. For similarity analysis the so called Key Point models are introduced, which extract the most essential features from the compressed information models as key points. An algorithm for similarity analysis, based on a pair-wise comparison of the key points from a pair of two key point models is given in the paper. The proposed technology for compression and similarity analysis is illustrated on the example of classifying 10 images of flowers into 5 different groups.
