Pattern segmentation in a binary/analog world: unsupervised learning versus memory storing

摘要: Abstract We discuss the problem of segmentation in pattern recognition. adopt model and general approach landmark paper by Wang, Buhmann von der Malsburg (Neural Computation, (1990), 2, 94–106), expand their a number ways. review solution to associative memory, which consists feature binding being expressed synchrony relations between oscillators or populations neurons. extend introducing law synaptic change, allows network learn structuring itself response stimuli with relevant features. interference completion learning new memories. also propose form multiplexing input information taking advantage time-structure neurons' response. It is based on assessment analog as well binary properties provides for an enhancement network's processing capacity. The relevance results biological systems pointed out.