Reinforcement Learning & Its Supervised View

Abstract

The task of training an artificial agent with test examples before placing it in the real environment is a typical learning methodology however making an agent learn by placing it directly in the real environment without providing any prior training and information about the environment and with no proper supervision is quite a difficult and challenging task. Such task is achieved by a different learning strategy called reinforcement learning. While the trained agents are based on the concept of supervised learning, the untrained agents form a part of special class of problems called exploration problems and which are studied under the domain of reinforcement learning. This paper gives the basic introduces the reinforcement learning, its types and focuses on the situation where reinforcement learning may be viewed under the domain of supervised learning.