ML-Terms used in Reinforcement Learning

In Reinforcement Learning, what is the definition of an "episode"?

• The number of steps the agent takes in the environment

• A sequence of actions taken by the agent

• A complete interaction cycle from the initial state to the terminal state

• The set of states the agent can be in during the learning process

What does the term "reward function" represent in Reinforcement Learning?

• A mathematical function that estimates the optimal policy

• The measure of how good the agent's policy is

• The set of actions the agent can choose from

• The function that calculates the immediate feedback for each action

What is the purpose of the "value function" in Reinforcement Learning?

• To compute the cumulative rewards for each action

• To measure the uncertainty of the agent's actions

• To estimate the expected long-term return from a given state

• To calculate the difference between the expected and actual rewards

In the context of Reinforcement Learning, what is "exploration"?

• The process of learning from a teacher's demonstrations

• The strategy of selecting known actions with the highest rewards

• The process of interacting with the environment and trying new actions

• The measure of the agent's performance compared to a baseline

What does the term "exploitation" refer to in Reinforcement Learning?

• The process of maximizing the cumulative rewards

• The strategy of selecting actions with the highest estimated value

• The measurement of how well the agent is performing

• The process of updating the agent's policy

In the context of Reinforcement Learning, what does "discount factor" represent?

• The difference between the expected and actual rewards

• The probability of taking a certain action in the environment

• The measure of how good the agent's decisions are

• The importance given to future rewards compared to immediate rewards

What is the "action space" in Reinforcement Learning?

• The set of all possible actions an agent can take in the environment

• The sequence of states the agent traverses during learning

• The process of adjusting the agent's policy to improve performance

• The set of training data samples used for learning

In Reinforcement Learning, what is the difference between an "episode" and a "trajectory"?

• They are different terms for the same concept.

• An episode refers to the sequence of states, while a trajectory refers to the sequence of actions.

• An episode refers to the sequence of actions, while a trajectory refers to the sequence of states.

• An episode is longer than a trajectory.

What is the role of the "baseline" in the advantage function used in Reinforcement Learning algorithms?

• To determine the initial state of the agent

• To estimate the average reward of an action in a given state

• To calculate the cumulative rewards over multiple episodes

• To represent the value of the optimal policy

What does the term "on-policy learning" mean in Reinforcement Learning?

• The agent learns from its past experiences

• The agent learns from a teacher or expert's demonstrations

• The agent learns from a different policy than the one it follows

• The agent learns from its own interactions with the environment using its current policy