Artificial Intelligence - Important Questions

Attempt each question before opening the matching module's revision section.

Module 1

1. Define AI using the rational-agent approach.
2. Specify and classify PEAS for a campus delivery robot.
3. Compare reflex, goal-based, utility-based, and learning agents.
4. Explain why rationality does not imply omniscience.
5. Classify a task environment with reasons.

Module 2

1. Formulate route finding as a search problem.
2. Compare BFS, UCS, DFS, greedy search, and A*.
3. Trace A* on a small weighted graph.
4. Solve a minimax tree with alpha-beta pruning.
5. Explain CSP backtracking, MRV, forward checking, and AC-3.

Module 3

1. Compare logic, semantic networks, frames, and scripts.
2. Represent a university domain with a semantic network.
3. Explain inheritance and exception handling in frames.
4. Demonstrate propositional resolution.
5. Distinguish an ontology from a knowledge graph.

Module 4

1. Solve a diagnostic problem using Bayes' rule.
2. Distinguish independence from conditional independence.
3. Explain Bayesian-network factorization.
4. Derive the Naive Bayes classification rule.
5. Explain HMM filtering, smoothing, and decoding.

Module 5

1. Draw and explain an artificial neuron.
2. Apply one perceptron weight update.
3. Explain why a single perceptron cannot solve XOR.
4. Compare sigmoid, tanh, ReLU, and softmax.
5. Explain forward propagation, backpropagation, and gradient descent.

Answer-writing checklist

• Define every important term.
• State the algorithm or formula before using it.
• Show frontier, probability, or weight calculations in a table.
• Draw a labelled diagram where appropriate.
• State assumptions, guarantees, and limitations.

PDF-focused numerical and diagram practice

1. Draw a PEAS table and classify a task environment under all standard properties.
2. Trace BFS, UCS, greedy, and A* on the same weighted graph and compare their selected nodes.
3. Solve a CSP using backtracking, MRV, forward checking, and one AC-3 pass.
4. Solve a minimax tree and mark every alpha-beta cutoff.
5. Convert a propositional knowledge base to CNF and perform resolution.
6. Test independence using both P(A|B)=P(A) and P(A and B)=P(A)P(B).
7. Perform one HMM filtering or Viterbi calculation from a transition/emission table.
8. Perform a perceptron update and one gradient-descent/backpropagation update.

For every algorithm answer, include inputs, steps, output, complexity or guarantee where applicable, and one limitation.