DSN RL SESSION 3

What is the main purpose of a Q-table in Q-learning?

• Store the value of state-action pairs

• Store rewards

• Store policy gradients

• Represent the environment

What does the parameter γ (gamma) represent in Q-learning?

• Exploration rate

• Discount factor for future rewards

• Learning rate

• Step size

What happens when ε (exploration rate) is set to 0?

• The agent explores more

• The agent always exploits

• The agent learns faster

• The agent resets Q-table

Which of these is a model-free learning algorithm?

• Q-learning

• Value Iteration

• Policy Iteration

• Dynamic Programming

Which library was used for environment simulation in the code?

• TensorFlow

• Gymnasium

• NumPy

• Keras

What does 'is_slippery=False' do in FrozenLake environment?

• Makes the lake more dangerous

• Increases the reward

• Makes movements deterministic

• Randomizes agent behavior

In the animation, what does each frame represent?

• A change in Q-table

• The agent’s current visual state

• A change in learning rate

• A random plot

What causes the Q-learning agent to stop an episode?

• When the Q-table is full

• When the episode is terminated or truncated

• When epsilon becomes zero

• When max reward is reached

Why is the epsilon decay used?

• To reduce rewards

• To reset Q-values

• To slowly move from exploration to exploitation

• To increase Q-table size

Which statement is true about model-free learning?

• It requires knowledge of environment dynamics

• It avoids trial-and-error

• It learns through interaction

• It doesn't use rewards