Terminology

Important Terms used in Reinforcement Learning:

1. Agent

   Agent is the "decision-maker" while solving a sequential decision making problem. It can also be called "learner" i.e. the one who learns to make sequential decisions.

2. Environment

   The world with which the agent interacts.

3. State (s)

   The current situation of the agent/environment.

4. Action (a)

   The move the agent chooses.

5. Reward (r)

   Feedback signal for the agent’s action.

6. Policy (\pi)

   The strategy the agent uses to decide actions.

7. Trajectory

   A sequence of states, actions, and rewards.

Interaction Cycle

Loop: Observe state \(s_t\) Choose action (a_t) based on policy π(s_t) Receive reward (r_t) and next state (s_{t+1}) from the environment Update policy to improve future decisions

MDP