Channel selection through a recommender system

Spectrum sharing between a cognitive network and a coexisting interfering network is considered. Among a collection of channels, in each time slot, each node in the cognitive network selects one channel based upon partial knowledge of the interference levels over these channels. The process of channel selection is treated as a recommender system, due to the similarity of recommending multiple items for multiple buyers and selecting multiple channels for multiple nodes. A modified collaborative filtering algorithm is proposed to perform prediction for the interference levels which would be suffered at receive nodes. Numerical experiments are presented to reveal the behavior of the proposed algorithm.