Appendix

A full workflow of channel number search with the proposed transitionary APS is shown in Algorithm 1. The overall procedure consists of two stages. To prove Theorem 3.1, we first show the case of two candidate decision Finally, to prove Theorem 3.1, we only need to extend Lemma 1 to the case of multiple candidate decisions. Here we cover more details in the design. We adopt policy gradient to maximize the reward function.