AML Reinforcement Learning & Decision Making

Question 1

In reinforcement learning, what does the agent seek to maximize over time?

Accepted Answer

Cumulative discounted reward

Answer

Prediction accuracy of the next state

Answer

Number of actions taken per episode

Answer

Shortest path to a terminal state

Question 2

Which of the following is NOT a standard component of a Markov Decision Process (MDP)?

Accepted Answer

Neural network architecture

Answer

State space

Answer

Transition probability function

Answer

Reward function

Question 3

What effect does setting the discount factor (γ) close to 0 have on a reinforcement learning agent?

Accepted Answer

The agent becomes myopic, prioritizing only immediate rewards

Answer

The agent learns faster due to stronger gradients

Answer

The agent explores more aggressively

Answer

The agent treats all future rewards equally to immediate ones

Question 4

What is the key distinction between model-based and model-free reinforcement learning?

Accepted Answer

Model-based RL learns or uses an environment model for planning; model-free learns directly from interactions

Answer

Model-based RL uses deep neural networks; model-free uses tabular methods only

Answer

Model-free RL requires more labeled data than model-based approaches

Answer

Model-based RL only works in discrete action spaces

Question 5

What is the exploration-exploitation dilemma in reinforcement learning?

Accepted Answer

Balancing trying new actions to gather information against using known high-reward actions

Answer

Choosing between GPU and CPU for training

Answer

Selecting between policy gradient and value-based methods

Answer

Allocating compute between training and inference phases

Question 6

What does a stochastic policy output for a given state in reinforcement learning?

Accepted Answer

A probability distribution over all possible actions

Answer

The single action with the highest Q-value

Answer

A random action drawn uniformly from the action space

Answer

The expected cumulative reward for that state

Question 7

What does the Bellman equation fundamentally express in reinforcement learning?

Accepted Answer

The recursive relationship between a state's value and the values of its successor states

Answer

The gradient of the policy loss with respect to network parameters

Answer

The backpropagation update rule for deep Q-networks

Answer

The optimal epsilon decay schedule for exploration