BE Computer Engineering (Pokhara University) Artificial Intelligence (PU, CMP 346) Question Paper 2079

(a) Define an intelligent agent and explain the PEAS (Performance measure, Environment, Actuators, Sensors) framework used to specify a task environment. Illustrate the PEAS description for an automated taxi-driving agent. (8)

(b) Classify task environments along the dimensions: fully observable vs. partially observable, deterministic vs. stochastic, episodic vs. sequential, static vs. dynamic, and discrete vs. continuous. For each pair, give one example. (6)

(a) Explain the A* search algorithm. Define the evaluation function f(n) = g(n) + h(n) and state the conditions of admissibility and consistency of a heuristic. Prove that A* with an admissible heuristic on a tree search is optimal. (10)

(b) Consider a graph where the heuristic and step costs are given below. Trace the order in which A* expands the nodes from start node S to goal node G, showing the f-values at each step:

Heuristics: h(S)=6, h(A)=4, h(B)=2, h(C)=2, h(G)=0. State the final path found and its total cost. (6)

(a) Explain the resolution refutation procedure in propositional/first-order logic. List the steps to convert a first-order logic sentence into Conjunctive Normal Form (CNF). (7)

(b) Given the following statements:

• All dogs are animals.
• All animals breathe.
• Tommy is a dog.

Represent these in first-order predicate logic, convert them to clausal form, and use resolution to prove that "Tommy breathes." (7)

(a) Draw the structure of a single artificial neuron (perceptron) and explain the role of weights, bias, summation and activation function. Compare the sigmoid, ReLU and step activation functions. (7)

(b) Explain the backpropagation algorithm used to train a multilayer feed-forward neural network. State why a non-linear, differentiable activation function is required for backpropagation to work. (5)

Compare Breadth-First Search (BFS) and Depth-First Search (DFS) on the basis of completeness, optimality, time complexity and space complexity. State one situation where DFS is preferred over BFS.

What is an expert system? Draw the architecture of a typical rule-based expert system and briefly explain the function of the knowledge base, inference engine and user interface. Differentiate between forward chaining and backward chaining.

Explain semantic networks and frames as techniques of knowledge representation. Represent the fact "A penguin is a bird that cannot fly but can swim" using a semantic network.

Differentiate between supervised, unsupervised and reinforcement learning with one suitable application example of each. What is meant by overfitting in a learning model and how can it be reduced?

Explain the hill-climbing search technique. Discuss the problems of local maxima, plateau and ridges encountered in hill climbing, and state one method to overcome each.

List and briefly explain the major stages of Natural Language Processing (NLP): morphological analysis, syntactic analysis, semantic analysis and pragmatic analysis. Why is ambiguity a central challenge in NLP?

State the rules of inference: Modus Ponens, Modus Tollens and Universal Instantiation. Using suitable propositional examples, demonstrate the application of Modus Ponens and Modus Tollens.

Write short notes on any TWO of the following: (a) Turing Test (b) Means-Ends Analysis (c) Adversarial search and the Minimax algorithm (d) Fuzzy logic