Graph convolution network deep reinforcement learning approach based on manifold regularization in cognitive radio network

摘要

In this paper, we propose GCNMR-ELM policy model for deep reinforcement learning approach in cognitive radio network and applications this policy model in cognitive radio on steelworks scene. This policy model combines the advantage of GCNMR framework and ELM algorithm. The aim is to enhance the data rate of spectrum sharing in cognitive?radio of steelworks, and Reduced policy model training time. The proposed policy model has a higher data rates in CR network can be provided; the convergence rate GCNMR-ELM policy model are faster than other policy model in the same number of iterations and GCNMR-ELM no increase in algorithm complexity. We provides extensive experiments on three different policy model in order to evaluate the performance of the proposed policy model. Experimental results show that our strategy model can effectively reduce the training time and provide higher data rate.