Machine Learning with Apache Spark Quick Start Guide

Reinforced learning

In reinforced learning, a reward (or punishment) system is employed to impact behavior over time, based on interactions between an agent and its wider environment. An agent will receive state information from the environment and, based on that state, will perform an action. As a result of that action, the environment will transition to a new state that is then provided back to the agent, typically with a reward (or punishment). The goal of the agent is to therefore maximize the cumulative reward it receives. For example, consider the case of a child learning good behavior from bad behavior and being rewarded for good behavior with a treat from its parents. In the case of machines, consider the example of computer-based ...