The emerge of Internet of Things (IoT) brings up revolutionary changes to the field of wireless communications. Providing connection to billions of IoT devices together with challenging IoT environments lead to the need of new technologies in order to fulfill IoT spectrum demands. Cognitive radio (CR) technology can be seen as one of the prominent solutions to the spectrum scarcity issue in IoT, where multiband cooperative spectrum sensing (CSS) is the key. In this thesis, we consider heterogeneous distributed learning-based CR networks (CRNs). Focus is given to learning approaches named as incremental, consensus, and diffusion. We perform a comparative study to investigate which one of them best fits cognitive IoT demands. Simulation results are performed illustrating potentials of diffusion and consensus algorithms and disadvantages of incremental for cognitive IoT systems. Lack of centralized control, increase in number of devices, and dynamic environments place a room for lots of challenges in the CSS process. Conventional CSS techniques have to be improved in order to fulfill sophisticated IoT requirements. One of the main challenges is cooperative secondary users' (SUs') scheduling to sense a subset of channels. To overcome the aforementioned challenge, in this thesis, we propose a novel multi-band CSS scheme, named as heterogeneous multi-band multiuser CSS (HM2CSS). The proposed scheme allows heterogeneous SUs to sense multiple channels and works by selecting cooperative SUs in two stages. Only SUs owning different information about channels are chosen to be cooperative. This is done by selecting leaders for each channel in the first stage and corresponding cooperative SUs in the second stage. Careful choice of leaders reflects the selection of cooperative SUs and hence, improves system performance. Then, diffusion learning algorithm is used to exchange locally sensed information among cooperative SUs for all channels. Further, the decision on the availability of channels is made. Extensive simulation results illustrate that the proposed HM2CSS scheme satisfies the IEEE 802.22 detection performance standard. Detection performance and CRN To my beloved family v Table of Contents vii List of Tables x List of Figures xi Nomenclature xiv List of Tables 4.1 Comparison of incremental, consensus, and diffusion-based learning approaches' characteristics. .