---

In spectrum sharing among different users, it is very essential to deal with the jammers in the network. This work is more important when the jammer is intelligent and capable of learning the communication patterns between users. In this paper, a spectrum sharing method has been suggested to deal with the jammers of the network. The proposed method is based on multi agent reinforcement learning. In the first stage, the interfered sub-bands are obtained by the base station. In the next stage, deep reinforcement learning is designed to offer each user a safe and appropriate spectrum. For this purpose, the effect of the jammer in the sub-band is obtained from the correlation between the transmitted pilot and the received feedback in each frequency interval. Then, the discount factor of the reinforcement learning algorithm is adjusted adaptively based on the correlation value. In this way, the sub-bands that are less affected by the jammer are assigned to the users. The proposed method has been evaluated in different scenarios and the results indicate that the sum rate of network converges to the desired level faster and the proposed method shows resistance against the jammer.