This document summarizes a B.Tech project thesis defense on distributed learning and adaptation in cognitive radio. The project aimed to develop a decentralized learning system to map secondary users to idle primary network channels without collision. The methodology used a novel Distributed Learning and Adaptive algorithm (DALA) policy based on stochastic learning and an Upper Confidence Bound algorithm. Simulation results showed the DALA policy converged channel selection probabilities and secondary users were assigned different channels. Mini-batch stochastic learning improved performance over DALA by reducing regret. Thresholding channel probabilities provided faster convergence. Future work may include variable step sizes, dynamic channel availability, and new ranking schemes.