BSCS Bachelor of Science in Computer Science: Artificial Intelligence Fundamentals 5 — Questions and Answers
Question 1: What is the purpose of the min-max theorem in game tree search for two-player zero-sum games?
- To maximize the combined utility of both players simultaneously
- To allow one player to maximize utility while the other minimizes it, determining the optimal strategy (Correct answer)
- To find a Nash equilibrium in cooperative games
- To prune branches where both players gain positive utility
Correct answer: To allow one player to maximize utility while the other minimizes it, determining the optimal strategy
Minimax alternates between maximizing (MAX player) and minimizing (MIN player) nodes to find each player's optimal strategy assuming rational opponents.
Question 2: In clustering, what does the silhouette score measure?
- The number of iterations needed for k-means to converge
- How well each data point fits its assigned cluster relative to other clusters (Correct answer)
- The total within-cluster sum of squared distances
- The ratio of between-cluster variance to within-cluster variance
Correct answer: How well each data point fits its assigned cluster relative to other clusters
The silhouette score measures how similar a point is to its own cluster versus neighboring clusters, ranging from -1 (wrong cluster) to +1 (well-clustered).
Question 3: What is the difference between model-based and model-free reinforcement learning?
- Model-based uses neural networks; model-free uses tabular methods
- Model-based learns a transition model of the environment; model-free learns a policy or value function directly from experience (Correct answer)
- Model-based requires continuous action spaces; model-free works only with discrete actions
- Model-based updates policies offline; model-free requires online interaction
Correct answer: Model-based learns a transition model of the environment; model-free learns a policy or value function directly from experience
Model-based RL builds an internal model of environment dynamics for planning, while model-free RL learns value/policy directly from sampled interactions.
Question 4: Which AI planning approach uses logic-based theorem proving to verify that a sequence of actions achieves a goal?
- Situation calculus with resolution theorem proving (Correct answer)
- Monte Carlo simulation
- Q-learning with a reward function
- Genetic algorithm fitness evaluation
Correct answer: Situation calculus with resolution theorem proving
Situation calculus represents actions and their effects in first-order logic, using resolution to prove that goal conditions hold after a plan is executed.
Question 5: What problem does batch normalization solve in deep neural network training?
- Reduces model size by compressing weight matrices
- Stabilizes training by normalizing layer inputs, reducing internal covariate shift (Correct answer)
- Increases training speed by skipping gradient computation on frozen layers
- Prevents overfitting by randomly dropping connections during forward passes
Correct answer: Stabilizes training by normalizing layer inputs, reducing internal covariate shift
Batch normalization normalizes activations within a mini-batch so each layer receives inputs with stable mean and variance, accelerating convergence.
Question 6: In propositional logic used in AI knowledge bases, what does the resolution inference rule do?
- Converts any formula to conjunctive normal form (CNF)
- Combines two clauses containing complementary literals to produce a new clause (Correct answer)
- Assigns truth values to all variables to satisfy a knowledge base
- Reduces the search space by pruning unsatisfiable branches
Correct answer: Combines two clauses containing complementary literals to produce a new clause
Resolution takes two clauses with complementary literals (P and ¬P) and derives a new clause with those literals removed.
Question 7: Which dimensionality reduction technique finds principal components by maximizing variance in the projected data?
- t-SNE (t-distributed Stochastic Neighbor Embedding)
- Principal Component Analysis (PCA) (Correct answer)
- Linear Discriminant Analysis (LDA)
- Autoencoders
Correct answer: Principal Component Analysis (PCA)
PCA finds orthogonal axes (principal components) in the direction of maximum variance via eigenvalue decomposition of the covariance matrix.
BSCS Bachelor of Science in Computer Science: Artificial Intelligence Fundamentals 5
What is the purpose of the min-max theorem in game tree search for two-player zero-sum games?