In this paper, we introduce an imagine network that can simulate itself through artificial association networks. Association, deduction, and memory networks are learned, and a network is created by combining the discriminator and reinforcement learning models. This model can learn various datasets or data samples generated in environments and generate new data samples.