An Improved Agent Strategy Training Method Based on DIAL
Runzhi Liu, Baolong Guo, Qianying Li, Biyan Wang · Journal of Physics Conference Series · 2019
Abstract The traditional multi-agent method has a slow convergence speed and needs to train many parameter groups. If no algorithm optimization is performed, it takes a lot of time to reach a large number in order to achieve optimal results in the whole process. We optimize the parameter settings in the traditional algorithm training process, accelerate the convergence speed of the intelligent body in learning, and achieve the convergence effect faster.