BSCS Bachelor of Science in Computer Science: Artificial Intelligence Fundamentals 2 โ Questions and Answers
Question 1: Which search algorithm uses a heuristic function h(n) to estimate the cost from node n to the goal and expands the node with the lowest f(n) = g(n) + h(n)?
- Dijkstra's algorithm
- A* search (Correct answer)
- Breadth-first search
- Iterative deepening DFS
Correct answer: A* search
A* search combines the actual path cost g(n) with a heuristic estimate h(n) to find the optimal path efficiently.
Question 2: In a Bayesian network, what does a conditional probability table (CPT) specify?
- The prior probability of the root nodes only
- The probability of a node given all possible combinations of its parents' values (Correct answer)
- The joint probability of all variables in the network
- The marginal probability of each variable independently
Correct answer: The probability of a node given all possible combinations of its parents' values
A CPT for a node lists P(node | parent combinations), capturing all conditional dependencies.
Question 3: What is the primary purpose of the softmax function in a neural network's output layer for multi-class classification?
- Introduce non-linearity in hidden layers
- Normalize outputs into a probability distribution summing to 1 (Correct answer)
- Prevent vanishing gradients
- Reduce overfitting by dropping neurons
Correct answer: Normalize outputs into a probability distribution summing to 1
Softmax exponentiates each output and divides by the sum, converting raw scores into class probabilities.
Question 4: Which phenomenon occurs when a decision tree is trained to perfectly fit training data, including noise, resulting in poor generalization?
- Underfitting
- Overfitting (Correct answer)
- Regularization
- Bagging
Correct answer: Overfitting
Overfitting happens when a model captures noise and random fluctuations in training data rather than the true underlying pattern.
Question 5: In reinforcement learning, what is the Bellman equation used for?
- Computing the entropy of a policy
- Expressing the value of a state recursively in terms of immediate reward and future value (Correct answer)
- Calculating the gradient of the loss function
- Normalizing the reward signal across episodes
Correct answer: Expressing the value of a state recursively in terms of immediate reward and future value
The Bellman equation decomposes V(s) = R + ฮณยทV(s'), forming the foundation for dynamic programming in RL.
Question 6: Which type of AI agent selects actions by comparing expected utility values and choosing the action that maximizes expected utility?
- Reflex agent
- Goal-based agent
- Utility-based agent (Correct answer)
- Learning agent
Correct answer: Utility-based agent
A utility-based agent uses a utility function to measure desirability of states and picks the action maximizing expected utility.
Question 7: What is the key characteristic of a PDDL (Planning Domain Definition Language) planning problem?
- It describes agent behavior using neural weights
- It defines initial state, goal state, and actions with preconditions and effects (Correct answer)
- It specifies reward functions for reinforcement learning
- It encodes probability distributions over state transitions
Correct answer: It defines initial state, goal state, and actions with preconditions and effects
PDDL separates domain (actions, predicates) from problem (initial state, goal), enabling classical planning algorithms.
BSCS Bachelor of Science in Computer Science: Artificial Intelligence Fundamentals 2
Which search algorithm uses a heuristic function h(n) to estimate the cost from node n to the goal and expands the node with the lowest f(n) = g(n) + h(n)?