Hands-On Reinforcement Learning for Games

Choosing a framework

As you may have surmised by now, writing your own RL algorithms and functions on top of a deep learning framework, such as PyTorch, is not trivial. It is also important to remember that the algorithms in this book go back about 30 years over the development of RL. That means that any serious new advances in RL take substantial effort and time—yes, for both development and especially training. Unless you have the time, resources, and incentive for developing your own framework, then it is highly recommended to graduate using a mature framework. However, there is an ever-increasing number of new and comparable frameworks out there, so you may find that you are unable to choose just one. Until one of these frameworks achieves ...