The model-driven power allocation (PA) algorithms in the wireless cellular
networks with interfering multiple-access channel (IMAC) have been investigated
for decades. Nowadays, the data-driven model-free machine learning-based
approaches are rapidly developed in this field, and among them the deep
reinforcement learning (DRL) is proved to be of great promising potential.
Different from supervised learning, the DRL takes advantages of exploration and
exploitation to maximize the objective function under certain constraints. In
our paper, we propose a two-step training framework. First, with the off-line
learning in simulated environment, a deep Q network (DQN) is trained with deep
Q learning (DQL) algorithm, which is well-designed to be in consistent with
this PA issue. Second, the DQN will be further fine-tuned with real data in
on-line training procedure. The simulation results show that the proposed DQN
achieves the highest averaged sum-rate, comparing to the ones with present DQL
training. With different user densities, our DQN outperforms benchmark
algorithms and thus a good generalization ability is verified.

本文研究了基于模型驱动的电力分配算法在具有干扰多接入信道（IMAC）的无线蜂窝网络中的应用。通过深度强化学习（DRL）的两步训练框架，采用深度 Q 网络（DQN）和深度 Q 学习算法（DQL）从离线学习中获得指定数据集，而对于在线学习过程中的真实数据来说，DQN 可进一步微调。这种方法比现有的 DQL 训练方法更好，具有很好的推广能力。