Reinforcement Learning CT-AI Practice Questions (Page 5)

Questions 21–24

1. 21application · easy

   An RL agent for a stock trading system uses the current portfolio value and market indicators as its state. Actions are buy, sell, or hold. The reward is the profit after each trade. What is the primary purpose of the value function in this system?

   • ATo calculate the immediate profit of the last trade.
   • BTo decide whether to buy, sell, or hold in each state.
   • CTo store the history of all past trades.
   • DTo estimate the expected future profit from a given market state.
   Reveal Solution Discussion Select an answer first
2. 22application · medium

   A reinforcement learning agent controls a robot vacuum. The state is the robot's position and battery level. Actions include moving to adjacent rooms or returning to the dock. The reward is +10 for cleaning a dirty room, -1 per move, and -5 for running out of battery. What does the agent learn to maximize?

   • AThe total discounted reward over time, balancing cleaning and battery management.
   • BThe number of actions taken per episode.
   • CThe immediate reward of the next action only.
   • DThe probability of each state being visited.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   What does a value function estimate in reinforcement learning?

   • AThe expected cumulative future reward from a state or state-action pair.
   • BThe immediate reward for taking a specific action.
   • CThe probability of transitioning from one state to another.
   • DThe number of actions available in a given state.
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   An RL agent controls a heating system in a smart building. The reward is +1 for maintaining a comfortable temperature and -10 for exceeding a high energy budget. Early in training, the agent rarely exceeds the budget but also rarely achieves comfort. Which of the following is the most likely cause?

   • AThe agent is exploring too much and therefore not learning a good policy.
   • BThe environment is non-stationary, so the agent cannot learn.
   • CThe reward function is too sparse to learn anything.
   • DThe agent is over-exploiting a suboptimal policy that avoids the large negative reward.
   Reveal Solution Discussion Select an answer first