MI

Question 1

What is Supervised Learning?

Answer 1

Supervised learning is a machine learning approach that trains on labeled data, where each example has a known correct output or label.

Question 2

Describe the structure of a Decision Tree.

Answer 2

A Decision Tree has: internal nodes for feature tests, branches for outcomes, and leaves for final classifications.

Question 3

How is a Decision Tree constructed?

Answer 3

1. Start with all data at the root. 2. Select a feature that best splits the data (e.g., using Gini or Information Gain). 3. Partition data accordingly. 4. Recursively repeat until a stopping condition (max depth, min samples, etc.) is reached.

Question 4

Give an example of a Decision Tree classification.

Answer 4

Example: Predicting laptop purchases. If the person is a student, check budget. If budget > $1000, predict purchase.

Question 5

What is the k-Nearest Neighbor (k-NN) algorithm?

Answer 5

k-NN is an instance-based learning method that classifies (or regresses) a new point based on its closest k neighbors.

Question 6

Describe the k-NN algorithm.

Answer 6

1. Choose k (number of neighbors). 2. Compute distance (e.g., Euclidean) from the new point to all training points. 3. Select the k closest. 4. Combine results via majority vote (classification) or average (regression).

Question 7

What are key evaluation metrics in supervised learning?

Answer 7

Common metrics include: Accuracy (correct/total), Precision & Recall (especially for imbalanced data), and ROC AUC (trade-off between TPR and FPR).

Question 8

What is a Neural Network?

Answer 8

A Neural Network is composed of: an input layer (features), one or more hidden layers (transformation of inputs), and an output layer (final predictions).

Question 9

How does backpropagation train a Neural Network?

Answer 9

1. Forward pass to get predictions. 2. Compute loss by comparing to true label. 3. Use gradients (via backpropagation) to update weights. 4. Repeat until convergence.

Question 10

Where are neural networks commonly used?

Answer 10

Neural networks are used in image recognition, speech processing, and self-driving cars.

Question 11

What is Unsupervised Learning?

Answer 11

Unsupervised learning discovers patterns from unlabeled data, unlike supervised learning which relies on labeled examples.

Question 12

What is Clustering?

Answer 12

Clustering groups similar data points based on similarity (e.g., distance).

Question 13

How does k-Means clustering work?

Answer 13

1. Pick the number of clusters k. 2. Randomly choose k initial centers. 3. Assign each point to the nearest center. 4. Update centers based on assignments. 5. Repeat until centers stabilize.

Question 14

How does Expectation-Maximization clustering work?

Answer 14

It alternates: E-Step (assign probabilistic memberships to clusters) and M-Step (update cluster parameters based on those memberships).

Question 15

Where is clustering used?

Answer 15

Clustering is used in customer segmentation, text classification, and medical diagnosis.

Question 16

What is a Constraint Satisfaction Problem (CSP)?

Answer 16

A CSP involves finding values for a set of variables within their domains so that all constraints are satisfied.

Question 17

What are the key components of a CSP?

Answer 17

1. Variables (items to assign). 2. Domains (possible values). 3. Constraints (rules that restrict valid assignments).

Question 18

How can CSPs be solved?

Answer 18

1. Backtracking Search (systematically try assignments). 2. Constraint Propagation (reduce domains before/during search). 3. Local Search (iteratively refine a complete assignment).

Question 19

What is Generalized Arc Consistency (GAC) in CSPs?

Answer 19

GAC ensures that for every constraint, each value in a variable’s domain is consistent with some value in the domains of other variables in that constraint.

Question 20

How does the Generalized Arc Consistency (GAC) algorithm work?

Answer 20

1. Initialize a queue with all constraints. 2. Make a variable’s domain consistent for each constraint. 3. If values are removed, re-check related constraints. 4. Repeat until no more changes.

Question 21

Why is Generalized Arc Consistency (GAC) useful?

Answer 21

It prunes inconsistent values early, reducing the search space and preventing needless backtracking.

Question 22

What is a Bayesian Network?

Answer 22

A Bayesian Network is a directed acyclic graph whose nodes represent variables and edges indicate probabilistic dependencies.

Question 23

How does inference work in a Bayesian Network?

Answer 23

1. Exact Inference (e.g., variable elimination). 2. Approximate Inference (e.g., Gibbs sampling, MCMC).

Question 24

What are some applications of Bayesian Networks?

Answer 24

They are used for medical diagnosis, spam filtering, robotics, and decision support systems.

Question 25

What is a search problem in AI?

Answer 25

A search problem defines an initial state, a goal state, and actions to move between states. The task is to find a path to the goal (often with minimal cost).

Question 26

What are some types of search algorithms?

Answer 26

1. Uninformed (BFS, DFS, UCS). 2. Informed (Greedy, A*). 3. Local (Hill Climbing, Simulated Annealing).

Question 27

How does A* search work?

Answer 27

A* uses f(n) = g(n) + h(n), where g(n) is the path cost so far and h(n) is a heuristic estimating the cost to the goal.

Question 28

Where are search algorithms used?

Answer 28

Search algorithms are used in pathfinding, AI planning, game AI, and robotics.

Question 29

What is Planning in AI?

Answer 29

Planning is selecting and organizing actions to transform an initial state into a goal state.

Question 30

What are the main components of a planning problem?

Answer 30

It has states, actions, an initial state, a goal state, and a plan (sequence of actions).

Question 31

What is STRIPS in planning?

Answer 31

STRIPS is a formalism that represents actions with preconditions, add lists, and delete lists.

Question 32

What does the Delete-Relaxation technique do?

Answer 32

It ignores negative (delete) effects of actions to simplify planning, often used for heuristic generation.

Question 33

What is a Multi-Agent System (MAS)?

Answer 33

A MAS involves multiple autonomous agents that can interact cooperatively or competitively.

Question 34

What is the Minimax algorithm used for?

Answer 34

Minimax is used in adversarial games to minimize potential losses assuming an optimal opposing player.

Question 35

What is Alpha-Beta pruning in Minimax?

Answer 35

Alpha-Beta pruning cuts off branches in the Minimax tree that won't affect the final choice, speeding up the search.

Question 36

What is a Nash Equilibrium?

Answer 36

A strategy profile in which no player can benefit by unilaterally changing their strategy.

Question 37

What is Planning under Uncertainty?

Answer 37

Planning in environments where actions have uncertain or probabilistic outcomes.

Question 38

What is a Markov Decision Process (MDP)?

Answer 38

An MDP is defined by (S, A, T, R, γ), where T is a transition function, R a reward function, and γ a discount factor.

Question 39

What is the Bellman Equation used for?

Answer 39

The Bellman Equation iteratively updates value functions to find an optimal policy in an MDP.

Question 40

What is Reinforcement Learning (RL)?

Answer 40

RL is where an agent learns from trial-and-error in an environment, receiving rewards or penalties for actions.

Question 41

What is Q-Learning in RL?

Answer 41

A model-free RL method updating Q(s,a) via: Q(s,a) ← Q(s,a) + α [r + γ * max Q(s',a') – Q(s,a)].

Question 42

What is an example of Reinforcement Learning?

Answer 42

A self-driving car learns optimal driving strategies through repeated trials and feedback (rewards/penalties).

Question 43

How do agents in Multi-Agent Systems make decisions?

Answer 43

They may use game theory, reinforcement learning, or rule-based systems depending on cooperation or competition.

Question 44

What is the role of a utility function in AI?

Answer 44

It represents an agent’s preferences, guiding action selection to maximize expected utility.

Question 45

What is an example of using Nash Equilibrium in real life?

Answer 45

If one chain lowers its prices while the other keeps prices high, the lower‐priced chain will attract more customers and boost its profits (both do it == bad)

Question 46

How does Value Iteration work in MDPs?

Answer 46

Value Iteration applies the Bellman update repeatedly to each state until values converge.

Question 47

What is Policy Iteration in MDPs?

Answer 47

It alternates between policy evaluation (computing state values for a policy) and policy improvement until optimal.

Question 48

What is the Exploration-Exploitation trade-off in RL?

Answer 48

Balancing trying new actions (exploration) and exploiting known high-reward actions to maximize long-term rewards.

Question 49

What is a common application of MDPs?

Answer 49

Robotics, where a robot must plan and act under uncertain transitions.

Question 50

What are some real-world applications of Planning?

Answer 50

Planning is used in robotics, autonomous driving, logistics, and scheduling tasks.

Question 51

How does STRIPS represent actions?

Answer 51

STRIPS actions are defined by preconditions, add effects (what becomes true), and delete effects (what becomes false).

Question 52

What is a heuristic function in AI planning?

Answer 52

A function estimating the remaining cost to reach a goal, used to guide search algorithms like A*.

Question 53

What are some real-world applications of CSPs?

Answer 53

(Scheduling problems, Sudoku, Map Coloring, and Timetabling.)

Question 54

What are the main techniques to solve CSPs?

Answer 54

(1. Backtracking Search, 2. Constraint Propagation, 3. Generalized Arc Consistency (GAC), 4. Local Search.)

Question 55

How does Generalized Arc Consistency (GAC) work?

Answer 55

(GAC ensures that for every constraint, each value in a variable's domain has a consistent corresponding value in other domains. It extends Arc Consistency to higher-order constraints.)

Question 56

What are the key properties of a Bayesian Network?

Answer 56

(1. Models uncertainty using probability, 2. Encodes conditional dependencies, 3. Uses a Directed Acyclic Graph (DAG).)

Question 57

What is conditional independence in Bayesian Networks?

Answer 57

(A variable is independent of its non-descendants given its parents.)

Question 58

What is an example of a real-world search problem?

Answer 58

(Pathfinding in Google Maps, solving puzzles like the 15-Puzzle, or finding bugs in software.)

Question 59

What are some common uninformed search strategies?

Answer 59

(Breadth-First Search (BFS), Depth-First Search (DFS), Uniform Cost Search (UCS).)