Canonical neural nets based on logic nodes

Logic nodes are distinguished from those classically used in neural computing by the fact that they store required responses to their input patterns in addressable locations rather than as connection strengths. The paper deals with `weightless´ systems or `logic nodes´. It is shown that by choosing synchronicity or asynchronicity, the storage of binary or probabilistic values, and the structuring of networks into various `canonical´ shapes, the entire paradigm of neural computing is covered. Also it is suggested that a clear path to implementation and insight is made possible by this approach