OpenAI Gym
In this section, we will study the OpenAI Gym tool. We will go through the motivations behind its creation and its main elements, learning how to interact with them to properly train a reinforcement learning algorithm to tackle state-of-the-art benchmark problems. Finally, we will build a custom environment with the same set of standardized interfaces.
The role of shared standard benchmarks for machine learning algorithms is of paramount importance to measure performance and state-of-the-art improvements. While for supervised learning there have been many different examples since the early days of the discipline, the same is not true for the reinforcement learning field.
With the aim of fulfilling this need, in 2016, OpenAI released OpenAI Gym (https://gym.openai.com/). It was conceived to be to reinforcement learning what standardized datasets such as ImageNet and COCO are to supervised learning: a standard, shared context in which the performance of RL methods can...