(SEM VI) THEORY EXAMINATION 2021-22 ARTIFICIAL INTELLIGENCE

ARTIFICIAL INTELLIGENCE (KME-062)

B.Tech Semester VI – Theory Examination 2021–22 

ARTIFICIAL-INTELLIGENCE-KME062

Artificial Intelligence is a branch of computer science that focuses on designing systems capable of performing tasks that normally require human intelligence. These tasks include reasoning, problem solving, learning, perception, language understanding, and decision making. The aim of Artificial Intelligence is not only to simulate human behavior but also to create rational agents that can act intelligently in complex and uncertain environments.

The uploaded question paper clearly shows that the syllabus emphasizes search algorithms, knowledge representation, logic and reasoning, probabilistic reasoning, machine learning, planning, natural language processing, vision, and robotics. To score well, answers must be written in clear, continuous paragraphs, showing conceptual depth, logical flow, and correct use of AI terminology rather than memorized definitions.

SECTION A – BASIC CONCEPTS OF ARTIFICIAL INTELLIGENCE

(Based on Section A on Page-1 of the paper) 

ARTIFICIAL-INTELLIGENCE-KME062

An Iterative Improvement Algorithm is a class of optimization algorithms that start with an initial solution and repeatedly improve it by making small changes until no better solution is found. These algorithms do not maintain a search tree; instead, they focus on improving a single current state. Hill climbing is a well-known example of this approach, where the algorithm moves in the direction of increasing value until it reaches a local optimum.

The different approaches of Artificial Intelligence include acting humanly, thinking humanly, thinking rationally, and acting rationally. Modern AI systems mainly follow the rational agent approach, where the system selects actions that maximize its expected performance.

First Order Logic (FOL) is a powerful knowledge representation language that extends propositional logic by introducing variables, predicates, functions, and quantifiers. It allows representation of complex relationships among objects and is widely used in AI reasoning systems.

Resolution is a rule of inference used in propositional and first-order logic. It works by combining clauses containing complementary literals to produce new clauses, and it forms the basis of automated theorem proving.

A Hidden Markov Model (HMM) is a probabilistic model used to represent systems that evolve over time with hidden states. It is widely used in speech recognition, natural language processing, and bioinformatics.

Inductive learning is a form of machine learning in which general rules or models are learned from specific examples. The system observes training data and infers patterns that can be applied to unseen data.

Clustering is an unsupervised learning technique in which data objects are grouped into clusters such that objects within the same cluster are similar to each other, while objects in different clusters are dissimilar.

Maximum a Posteriori (MAP) estimation refers to selecting the hypothesis that has the highest posterior probability given the observed data. It combines prior knowledge with evidence from data.

Speech recognition applications include voice assistants, automated customer support systems, dictation software, and hands-free control systems.

Practical Natural Language Processing refers to computational techniques used to enable machines to understand, interpret, and generate human language in real-world applications such as translation, summarization, and sentiment analysis.

SECTION B – SEARCH, LOGIC, LEARNING & AGENTS

(Based on Section B on Page-1) 

ARTIFICIAL-INTELLIGENCE-KME062

Heuristic search is an informed search technique that uses domain-specific knowledge to guide the search toward a goal more efficiently. A good heuristic function estimates the cost from the current state to the goal. Desirable properties of a heuristic include admissibility, consistency, and efficiency, as these properties ensure optimal and fast solutions.

Propositional Logic is a formal language used to represent facts about the world using propositions that can be true or false. Inference rules such as Modus Ponens, Modus Tollens, Resolution, and And-Elimination allow new facts to be derived logically from known facts.

Bayesian Theory provides a mathematical framework for reasoning under uncertainty. In Bayesian classification, the prior probability represents initial belief about a hypothesis, while the posterior probability represents the updated belief after observing evidence. Bayes’ theorem connects these probabilities and forms the foundation of probabilistic learning.

Supervised learning and unsupervised learning differ in the availability of labeled data. Supervised learning uses labeled examples to train predictive models, whereas unsupervised learning discovers hidden patterns without predefined labels.

Intelligent agents are systems that perceive their environment through sensors and act upon it using actuators to achieve goals. Communicating agents can cooperate, coordinate, or negotiate with other agents to solve complex tasks.

SECTION C – SEARCH SPACE AND REASONING

(Based on Section C on Page-1) 

ARTIFICIAL-INTELLIGENCE-KME062

The Hill Climbing Algorithm suffers from several problems such as local maxima, plateaus, and ridges. These issues prevent the algorithm from reaching the global optimum, especially in complex search spaces.

A problem space represents all possible states of a problem along with transitions between them. Defining a problem as a state space search involves specifying the initial state, goal state, actions, and cost function. This formulation allows AI systems to apply general search algorithms to solve diverse problems.

LOGIC-BASED REASONING

(Based on Question 4 on Page-1) 

ARTIFICIAL-INTELLIGENCE-KME062

The Unification Algorithm is used in predicate logic to determine whether two logical expressions can be made identical by substituting variables. It plays a critical role in automated reasoning and logic programming.

Forward chaining and backward chaining are inference strategies used in rule-based systems. Forward chaining starts from known facts and applies rules to derive conclusions, while backward chaining starts from a goal and works backward to verify whether known facts support it.

PLANNING, INFERENCE AND LEARNING

(Based on Page-2 of the paper) 

ARTIFICIAL-INTELLIGENCE-KME062

Planning graphs are data structures used to represent planning problems efficiently. They help in identifying possible actions and states over time and are useful in real-world planning where actions must be executed under uncertainty.

Approximate inference in Bayesian networks is required when exact inference becomes computationally expensive. Techniques such as sampling, Monte Carlo methods, and particle filtering are commonly used.

Decision Trees are supervised learning models that represent decisions in a tree-like structure. They are easy to interpret and widely used in classification tasks.

Reinforcement Learning (RL) is a learning paradigm where an agent learns optimal behavior by interacting with an environment and receiving rewards or penalties. Applications include robotics, game playing, and autonomous systems.

ROBOTICS AND VISION

(Based on Question 7 on Page-2) 

ARTIFICIAL-INTELLIGENCE-KME062

The architecture and configuration of robots include sensors, actuators, control systems, and decision-making modules. Proper configuration allows robots to perform tasks autonomously.

Image processing operations such as enhancement, segmentation, and feature extraction enable vision systems to perceive the environment. Vision plays a crucial role in navigation and manipulation by allowing robots to identify objects and avoid obstacles.

HOW TO WRITE ARTIFICIAL INTELLIGENCE ANSWERS IN THE EXAM

In Artificial Intelligence, never write answers in short bullet points. Always begin with a clear definition, followed by explanation of concepts, examples, and significance. Use correct AI terminology such as heuristic search, inference, learning, probability, and agents. Examiners give maximum weightage to conceptual clarity, logical reasoning, and proper explanation of algorithms.