Setting a powerful spectrum sensing and access policy increases the throughput of cognitive radio networks (CRNs). In this paper, the problem of maximizing the average throughput of a CRN through setting proper sensing sequences is investigated. In addition, a systematic neural network-based optimization approach is developed which avoids challenges associated with the conventional analytical solutions. The proposed intelligent learning and optimization cycle, based on a cooperation between two kinds of well-known artificial neural networks, finds the optimal sensing sequence for each secondary user without any prior knowledge or presumptions about the wireless environment. The structure of the proposed scheme is discussed in detail, and its efficiencies are verified through a set of illustrative numerical results.