Logic theory of binary neural networks

We foresee the applicability of binary neural networks to a number of engineering problems. We feel that its logic theory should be studied so as to make it useful in engineering in the future. Past researchers have only actively studied computational ability of binary neural networks. Our study reveals that, as neural networks can be viewed as constraint satisfaction networks (CSNs), it needs an inherent logic theory consisting of two-state (excitatory state and inhibitory state) decisions, weak inference, rule types (excitatory and inhibitory), strength identity and contradiction (superior and inferior). Superior contradiction is discussed and defined. The process by which a neural network seeks a solution corresponds to elimination of the superior contradiction. the neural networks can be described as self-made, open, flexible, and self-adaptive CSNs.