Question 1

Q.1. (a) Discuss the different application areas of Artificial Intelligence. [8]

Accepted Answer

AI finds application across expert systems, natural language processing, computer vision, robotics, gaming, healthcare diagnostics, autonomous vehicles, and speech recognition, among many other domains; machine learning systems automatically learn patterns and decision rules from data without being explicitly programmed, whereas expert systems encode a human expert's explicit domain knowledge and inference rules directly, provided by knowledge engineers rather than learned from data.

Question 2

Q.1. (a) What are the characteristics of a good production system? [8]

Accepted Answer

A good production system should be sound (producing only valid/correct rule applications), complete (able to reach all achievable solution states), efficient in its search strategy, and should exhibit appropriate conflict-resolution behavior when multiple rules are simultaneously applicable; state space search represents a problem as a graph of states connected by operators (actions), with the search algorithm's task being to find a path from the initial state to a designated goal state through this state space.

Question 3

Q.2. (a) Discuss the characteristics of knowledge representation. Define prop…

Accepted Answer

Knowledge representation is characterized by properties such as representational adequacy, inferential adequacy, inferential efficiency, and acquisitional efficiency, with knowledge itself possessing properties including being declarative or procedural, structured, and having varying degrees of certainty; the hill climbing algorithm is a local search technique that iteratively moves from the current state to a neighboring state with a better evaluation-function value, terminating when no neighboring state offers further improvement.

Question 4

Q.2. (a) Discuss about the following terms in knowledge representation: (i) V…

Accepted Answer

In knowledge representation, validity refers to a logical statement being true under every possible interpretation, satisfiability refers to a statement being true under at least one interpretation, and contradiction refers to a statement being false under every possible interpretation; in a support vector machine, the positive margin is the distance from the decision boundary to the nearest positive-class support vector, and the negative margin is the corresponding distance to the nearest negative-class support vector, with the SVM training objective being to maximize the combined (total) margin between these two classes.

Question 5

Q.3. (a) Explain the concept of neural network. [8]

Accepted Answer

A neural network is a computational model loosely inspired by biological neurons, consisting of interconnected layers of simple processing units (neurons) that apply weighted sums and non-linear activation functions to transform input data into a desired output, capable of learning complex non-linear input-output mappings; the learning algorithm in neural networks, most commonly backpropagation combined with gradient descent, iteratively adjusts the network's connection weights to minimize the error between the network's actual output and the desired target output across the training dataset.

Question 6

Q.3. Explain the different characteristics of perception. Also discuss the ap…

Accepted Answer

Perception in AI is the process by which an intelligent system interprets sensory input (visual, auditory, or other sensor data) to construct a meaningful internal representation of its environment, characterized by properties including robustness to noise, invariance to irrelevant transformations, real-time processing capability, and multi-modal integration; perception finds application in computer vision, speech recognition, robotics navigation, autonomous vehicles, and human-computer interaction.

Question 7

Q.4. (a) What is machine learning systems? Explain. [8]

Accepted Answer

A machine learning system is a computational system that automatically improves its performance on a given task through experience (training data), without being explicitly programmed with a fixed set of task-specific rules; linear regression is a representative supervised learning algorithm that learns a linear relationship between input features and a continuous target output by minimizing the sum of squared prediction errors across the training data.

Question 8

Q.4. (a) Discuss about support vector machine. [8]

Accepted Answer

A Support Vector Machine (SVM) is a supervised learning algorithm that finds the maximum-margin hyperplane separating two classes of data, extendable to non-linearly separable data through the kernel trick, which implicitly maps data into a higher-dimensional space where linear separation becomes possible; supervised learning trains a model using labeled input-output example pairs to learn a mapping function, whereas unsupervised learning works with unlabeled data alone to discover inherent structure or groupings without any predefined target output.

Question 9

Q.5. What is fuzzy logic? Differentiate between predicate logic and fuzzy log…

Accepted Answer

Fuzzy logic is a form of many-valued logic that allows a proposition's truth value to range continuously between completely true (1) and completely false (0), rather than being restricted to the strictly binary true/false values of classical predicate logic, allowing it to naturally represent and reason with the inherent vagueness and imprecision present in much real-world knowledge and natural-language concepts; predicate logic uses crisp set membership and strictly bivalent truth values, whereas fuzzy logic uses fuzzy set membership functions and continuous-valued truth degrees.

Question 10

Q.5. (a) List different genetic algorithm approaches. [8]

Accepted Answer

Genetic algorithm approaches include the simple/canonical genetic algorithm, steady-state genetic algorithm, generational genetic algorithm, and various specialized variants such as messy genetic algorithms and parallel/island-model genetic algorithms, differing mainly in population replacement strategy and selection scheme; fuzzy set operations include union (maximum), intersection (minimum), and complement, while fuzzy quantifiers (absolute and relative) express fuzzy statements about quantity, such as 'most' or 'about five', extending classical logical quantifiers (for-all, there-exists) to accommodate graded, imprecise quantity expressions.

Artificial Intelligence Techniques