Multi-Agent Deep Reinforcement Learning Based Spectrum Allocation for D2D Underlay Communications | IEEE Journals & Magazine | IEEE Xplore