Reinforcement learning techniques

With this background in reinforcement learning, in the next few sections we are going to look at some of the formal techniques for exploration into the search space with the goal of maximizing the rewards in an optimal way. 

Markov decision processes

In order to understand the Markov decision processes (MDPs), let us define two environment types:

• A deterministic environment: In a deterministic environment, an action taken within a particular state of the environment determines a certain outcome. For example, in the game of chess out of all the possible moves at the beginning of the game, when we move a pawn from e4 to e5, the immediate next step is certain and does not differ across various games. There is also a level of certainty of reward in a deterministic environment along with the next possible state(s).
• A stochastic environment: In the case of a stochastic environment, there is always a level of randomness and uncertainty in terms of next state of the...