BE Computer Engineering (IOE, TU) Artificial Intelligence (IOE, CT 653) Question Paper 2079

(a) Define an intelligent agent. With the help of a block diagram, explain the structure of a learning agent, clearly describing the role of the performance element, learning element, critic, and problem generator. (7 marks)

(b) Consider an automated vacuum-cleaning robot operating in a house. Specify its task environment using the PEAS (Performance measure, Environment, Actuators, Sensors) framework, and classify the environment along the dimensions: observable vs. partially observable, deterministic vs. stochastic, episodic vs. sequential, static vs. dynamic. (5 marks)

(a) State and prove the conditions of admissibility and consistency of a heuristic function. Show that every consistent heuristic is admissible. (6 marks)

(b) Consider the following weighted graph where the start node is S and the goal node is G. The edge costs and the heuristic estimates h(n) to the goal are given below:

Edges (with cost): S-A=3, S-B=5, A-C=4, B-C=2, C-G=4, A-G=12.

Trace the A* search algorithm step by step, showing the contents of the open and closed lists at each iteration, and determine the optimal path from S to G and its total cost. (10 marks)

(a) Convert the following statements into well-formed formulae in first-order predicate logic:

1. Every student who studies hard passes the examination.
2. Some students do not like every subject.
3. All birds can fly except penguins. (6 marks)

(b) Given the knowledge base:

• All people who own a dog are animal lovers.
• Animal lovers do not kill animals.
• Either Jack or Curiosity killed the cat named Tuna.
• Jack owns a dog.

Convert the statements to clause form and use resolution refutation to prove that Curiosity killed the cat. (6 marks)

(a) Draw the architecture of a feed-forward multilayer perceptron and derive the weight-update rule for a single hidden layer using the backpropagation algorithm. Clearly state all the assumptions and the role of the activation function. (8 marks)

(b) Why is a single-layer perceptron unable to solve the XOR problem? Show, with a suitable diagram, how a multilayer network overcomes this limitation. (4 marks)

Compare Breadth-First Search (BFS) and Depth-First Search (DFS) in terms of completeness, optimality, time complexity, and space complexity. Under what circumstances would you prefer Iterative Deepening DFS over either of them?

Explain the hill-climbing search technique. What are the problems of local maxima, plateau, and ridges encountered in hill climbing, and how can each of them be addressed?

What is knowledge representation? Represent the following facts using a semantic network: Tweety is a bird, all birds are animals, birds can fly, and animals can breathe. State one advantage and one limitation of semantic networks.

With a neat block diagram, explain the architecture of an expert system. Differentiate between forward chaining and backward chaining inference mechanisms with a suitable example.

Differentiate between supervised and unsupervised learning with appropriate examples. Briefly describe how overfitting occurs in a machine learning model and list two techniques to reduce it.

List and briefly explain the different phases of Natural Language Processing (NLP) (lexical, syntactic, semantic, discourse, and pragmatic analysis). Why is ambiguity a fundamental challenge in NLP? Give one example.

Define unification in predicate logic. Find the Most General Unifier (MGU), if it exists, for each of the following pairs of expressions, or state why unification fails:

(a) P(x, f(y)) and P(a, f(g(z)))

(b) Q(x, x) and Q(a, b)

What is the Turing Test? Discuss whether passing the Turing Test is a sufficient condition for a machine to be considered truly intelligent. Mention one major criticism of the test.

Explain the Minimax algorithm for game playing. How does alpha-beta pruning improve its efficiency without affecting the final decision?