Petty little joy # 6: OpenAI Gym - play games and control robots
We are used to the fact that machine learning specialists spend most of their time working on analyzing tabular data or training a neural network to lay out photos of cats and dogs in two neat heaps. For this is the will of the business - most often you need to give forecasts, classify data and build models.
Today we will air our brains and knead synapses flabby with boredom with the help of something interesting - OpenAI Gym .
OpenAI Gym lets you hone your reinforcement learning skills. This is one of the types of machine learning tasks in which you are offered a certain environment with certain rules and an agent that can act in this environment. For example, a game about Mario is an environment, and an Italian plumber figure in it, controlled by a machine algorithm, is an agent.
OpenAI Gym offers users a choice of a healthy set of environments and agents - simulators of old computer games (including classic games from Atari), virtual robots, industrial machines and primitive environments with simple physical laws.
All these complex and diverse environments are presented in the form of understandable and standard numerical models, which is easy and simple to manage from Python scripts. Agents are also easily managed from within a Python application.
A great and comfortable environment to spend a few nights with a bored machine learning specialist or a sad Python developer.