A priority table of available channels in a network is structured by analyzing channel availability with cooperative spectrum sensing. Using this table, a channel selection learning algorithm based on the experience-weight attraction (EWA) game learning is proposed. Simulations are carried out to compare the learning automata-based algorithm with no-regret learning algorithm. The results show that, by learning historical experience, the algorithm can select channels with the best availability for cognitive users, increase the effective system throughput, and have better equity in resource allocation.