On the basis of the hypothesis that the existing relevant knowledge influences the formation of the new knowledge, this thesis sets up the maximum entropy fundamental model of the stochastic prosess of knowledge innovation by applying the probability & statistics theory, the optimization theory and the maximum entropy principle comprehensively, and puts forward the conditional probability P(y|x) of the new knowledge y∈Y which is restrained by the existing knowledge x∈X. This model posesses such character that the randomness and the causality are unity of opposites