In light of the deficiency of the original model of Probabilistic Logic Neuron (PLN)network, this paper presents a revised model and shows that its robustness and speed are improved. The authors also indicate that the original model of PLN network,the Hopfield's learning rule and the evolution rule of Boltzman machine are special cases of the new model. Finally, the application of the generalized PLN network to associative memory is given. The computer simulation results show that robustness and convergence speed of the new model are much better than the original one.