Dual encoding as a means of improving neural net performance

A technique for encoding complementary memories in auto-associative neural nets has been employed in order to improve memory recall performance. The mechanism of dual encoding is adapted. From ideas concerning interhemispheric communications and the functional asymmetries of the cerebral hemispheres. A pattern stored in one net can be regenerated from that in the other, but the patterns are complementary and not simple mirror-images. By allowing bilateral auto-associative neural nets to store and recall complementary information, a significant improvement in recall performance can be obtained. The mechanism of bilateral storage is, in principle, applicable to other auto- and hetero-associative neural nets.