this work proposes a new method to solve the unstable deep RL network, asynchronous methods, with higher performance and shorter training time. this advantage offers practical benefits. it has some experiments on combining different basic algorithm, and all of them get a higher performance, which proves that it can be generally applied on many RL algorithm.
this work proposes a new method to solve the unstable deep RL network, asynchronous methods, with higher performance and shorter training time. this advantage offers practical benefits. it has some experiments on combining different basic algorithm, and all of them get a higher performance, which proves that it can be generally applied on many RL algorithm.