Deep Reinforcement Learning-based Spectrum Assignment with Multi-metric Reward Function and Assignable Boundary Slot Mask
Masayuki Shimoda, Takafumi Tanaka · 26th Optoelectronics and Communications Conference · 2021
We propose a deep reinforcement learning-based spectrum assignment with a multi-metric reward function and assignable boundary slot mask. The mask restricts options to assignable ones and multi-metric function provides informative clues for training, outperforming first-fit.