REINFORCEMENT LEARNING

Reinforcement Learning is a technique for learning by making a series of repeated decisions. The aim of this technique is to discover a policy that produces a desired outcome. This policy is a recipe for how the agent should behave in different states of a dynamic system. In the case of reinforcement learning, the action of the agent is influenced by the reward or punishment it receives.

​

The goal of reinforcement learning is to maximize the number of rewards an agent receives after each good action. This process relies on a signal that is sent by the environment. It is local in nature, but long-term in effect. This means that actions in one state lead to actions in future states. Reinforcement learning has an optimization objective called the discounted cumulative reward signal.

​

In order to create a policy, an agent learns a model of the environment by sampling states and rewards. By analyzing the outcomes, it can then predict the expected reward in the future. It is this process that is known as policy iteration. By experimenting with different policies, it is possible to create a policy that is optimal for a given task.

​

Reinforcement learning methods provide a more dynamic learning environment than traditional machine learning methods. This method is constantly breaking new ground and continues to grow. Reinforcement Learning is an integral part of learning theory and is the basis of many practical applications. 

​

Reinforcement learning works best when the agent interacts with its environment. This allows the agent to build a repertoire of examples that it can use to make better decisions. This process is similar to that of the human brain. This method requires an agent to interact with an environment that is complex and high-dimensional.