Reinforcement Learning (AD-802 (B)) - Important Questions

Define reinforcement learning and give a real-life example.

Explain various approaches to implement reinforcement learning.

Differentiate between reinforcement learning and supervised learning.

What is the key objective of bandit algorithms in reinforcement learning?

After 12 iterations of the UCB1 algorithm applied on a 4-arm bandit problem, we have n1=3, n2=4, n3=3, n4=2 and Q12(1)=0.55, Q12(2)=0.63, Q12(3)=0.61, Q12(4)=0.40. Which arm should UCB1 play next? Justify with calculation.

What is meant by passive and active reinforcement learning and how do we compare the two?

Explain the following terms with respect to bandit algorithms: Median Elimination, Upper Confidence Bound (UCB) algorithm, and Probably Approximately Correct (PAC).

Explain the Q-Function and Q-Learning Algorithm.

Discuss Maximum Likelihood and Least Square Error Hypothesis.

Explain Fitted-Q and Deep Q-Learning Problems.

Explain any one advanced Q-learning algorithm.

Explain learning policies by imitating optimal controllers.

Explain inverse reinforcement learning.

Explain the following: (i) Deep Q-Network (DQN) and Policy Gradient (ii) Applications and issues in deep reinforcement learning.