AN ADAPTIVE COMPUTING SYSTEM FOR THE SYNTHESIS OF THRESHOLD LOGIC NETWORKS,

The report describes development of an adaptive program to examine real problems with unknown solutions improving performance on the basis of system experience. A new scheme, reflecting Boolean functions on a Veitch diagram, is developed for classification of Boolean functions. A theorem is presented for the calculation of Rademacher-Walsh coefficients of self-dual functions. A heuristic training algorithm is developed which evolves a minimal element threshold logic net. The value of making exploratory studies through a learning interactive program are illustrated and discussed.