Deep Learning with PyTorch Quick Start Guide

Reinforcement learning

Reinforcement learning is somewhat different to other methods and is often classified as an unsupervised method because the data it uses is not labeled in the supervised sense. Reinforcement learning probably comes closer to the way humans interact and learn from the world than other methods. In reinforcement learning, the learning system is called an agent and this agent interacts with an environment by observation and by performing actions. Each action results in either a reward or a penalty. The agent must develop a strategy or policy to maximize reward and minimize penalties over time. Reinforcement learning has applications in many domains, such as game theory and robotics where the algorithm must learn its environment ...