Recall in saturated associative neural networks

Summary form only given, as follows. An associative neural network was enhanced to provide both primary and secondary associative relationships on recall. Input patterns, which were conceptually related, established extra relationships in the modified associative network. Not only did this network recall the original associative pattern when presented with a given input, but it could identify the conceptually related input patterns. The output was separated into two groups. The primary output was the expected pattern associated with the input during training. The secondary output consisted of the patterns associated with other inputs. This secondary information was limited to output patterns which were conceptually related to the training input.<>