The Reinforcement Learning Competition 2014

R einforcement learning (RL) is the problem of learning how to act only from interaction and limited reinforcement. An agent takes actions in an unknown environment , observes their effects, and obtains rewards. The agent's aim is to learn how the environment works in order to maximize the total reward obtained during its lifetime. RL problems are quite general. They can subsume almost any artificial intelligence problem, through suitable choices of environment and reward. For example, learning to play games can be formalized as a problem where the environment is adversarial and there is a positive (negative) reward when the agent wins (loses). An introduction to RL is given by the book of Sutton and Barto (1998), while a good overview of recent algorithms is given by Szepesvári (2010). Most RL problems are formalized by modeling the interaction between the agent and the environment as a discrete-time process. At each time t, the agent takes an action at from some action set A and obtains an observation x t+1 ∊ X and a scalar reward r t+1 ∊ R. Part of the problem is estimating a model for the environment, that is, learning how the observations and rewards depend upon the actions taken. However , the agent's actual objective is to maximize the total reward n Reinforcement learning is one of the most general problems in artificial intelligence. It has been used to model problems in automated experiment design, control, economics, game playing, scheduling, and telecommunications. The aim of the reinforcement learning competition is to encourage the development of very general learning agents for arbitrary reinforcement learning problems and to provide a test bed for the unbiased evaluation of algorithms.