Introduction to Artificial Intelligence Quiz

Planning agents consider the future consequences of their actions and generate a sequence of actions to achieve a goal.

Uninformed search methods use no knowledge about the problem to guide their search.

A* search combines the cost of getting to a node (g(n)) with an estimated cost to reach the goal (h(n)) to decide the next node to explore.

Reflex agents use condition-action rules to respond to the current environment, without considering the future.

Problem-solving agents consider the consequences of their actions and try to achieve a specific goal.

Depth-first search explores a single branch of the search tree until it reaches a goal or a dead end, then backtracks and explores another branch.

Breadth-first search explores all nodes at a given depth before moving to the next level, ensuring the shortest path to the goal is found first.

Uniform-cost search expands the node with the lowest total path cost, prioritizing efficiency.

The initial belief about the probability of an event before any evidence is considered.

A function representing the probability distribution of a continuous random variable.

Two variables are independent if the probability of one event does not influence the probability of the other event.

When the probability of one variable is independent of another variable given a third variable.

Simplifying the probability of multiple events occurring by breaking it down into conditional probabilities.

Updating your beliefs about the probability of an event based on new evidence.

A formula that calculates the probability of an event (hypothesis) given that another event (evidence) has occurred.

A graphical model representing probabilistic relationships between variables.

Two events are independent if the probability of both events occurring is equal to the product of their individual probabilities.

The frequentist view of probability sees probability as the long-run frequency of events.

Kolmogorov's second axiom states that the probability of any event is non-negative.

A marginal distribution is calculated by summing the probabilities over all possible values of the other variables.

If two events are independent, the probability of both events occurring is the product of their individual probabilities.

In a probability distribution, the sum of all probabilities must equal 1.

The Principle of Maximum Expected Utility suggests that agents should choose actions that maximize their expected utility.

The horizon effect occurs when a search algorithm explores many similar states repeatedly, leading to inefficiency. Extending the search using quiescence search helps by exploring critical points in the search tree, preventing redundant exploration of similar states.

Quiescence search is a technique used in game AI to reduce the horizon effect by extending the search at critical points in the game tree. These points are typically associated with significant changes in the game's state, ensuring that the algorithm explores these changes more thoroughly.

AlphaGo is a state-of-the-art game AI developed by Google DeepMind, known for its proficiency in Go, a complex board game. It's a prime example of a game AI employing Monte Carlo Tree Search (MCTS).

AlphaGo Zero is a variant of AlphaGo that was trained entirely through self-play. It learned by playing against itself without any human data, showcasing the power of reinforcement learning.

Transposition tables are data structures used to store and retrieve previously explored game states reached through different move sequences. They help avoid redundant calculations and speed up search algorithms, especially in games with repetitive states.

In MCTS, exploitation refers to the strategy of selecting moves that are believed to be most promising based on past simulations. It prioritizes using knowledge gathered from previous trials to make optimal decisions.

Adversarial search is a technique used in areas like cybersecurity, game AI, and other domains where a decision-maker must anticipate and counter the actions of an opponent. It analyzes the actions of the opponent to choose a strategy that minimizes potential loss.

Deep Blue was a chess-playing computer system developed by IBM. It achieved significant milestones by defeating chess grandmaster Garry Kasparov in 1997, showcasing the power of AI in playing strategic games.

In a Fork structure, A ← B → C, A and C are conditionally independent given B. This means that knowing the value of B makes A and C independent of each other.

Approximate inference techniques, such as Monte Carlo methods, are used to estimate probabilities in complex Bayesian networks where exact calculations are computationally infeasible.

Bayesian Parameter Learning uses both observed data and prior knowledge to update beliefs about parameters. It combines prior beliefs with evidence to arrive at more accurate estimates.

A Bayesian Network structure like A → B ← C, where B is influenced by both A and C, is known as a collider. This means that B serves as a common cause for A and C.

Bayesian Networks are graphical models that represent probabilistic relationships between variables. They allow for reasoning with conditional dependencies, meaning that the probability of an event can be influenced by other events.

D-Separation is a technique used to determine conditional independence relationships between variables in a Bayesian Network. It helps us understand which nodes are independent given the knowledge of other nodes.

Bayesian Networks are powerful tools for modeling uncertain events and their relationships. They allow us to represent and reason about probabilities in a complex world.

Variable Elimination is a technique used to simplify complex Bayesian Networks by systematically eliminating variables, starting with those that are not directly related to the query variable. This reduces computational complexity and makes inference more efficient.

If A and C are independent given B, then knowing the value of B doesn't add any information about the relationship between A and C. In other words, A and C are independent, even if you're provided information about B.

Variable elimination systematically eliminates variables to simplify calculations, making inference efficient. It involves performing calculations on smaller sets of variables and combining the results.

A Bayesian Network with the structure A → B → C implies that A directly influences B, which in turn influences C. There is no direct influence from A to C, only through the intermediary B.

MAP estimation combines observed data with prior beliefs (previous knowledge) to find the most probable explanation for the data.

Approximate inference is used when the Bayesian Network is too large or complex for exact inference methods. It provides an estimated solution instead of an exact one.

Bayesian Networks are well-suited for representing uncertain knowledge with probabilistic dependencies between variables. They allow us to model and reason about relationships between uncertain events.

d-Separation in Bayesian Networks relies on blocking paths between nodes. It determines conditional independence by checking if there is an active or blocked path between two variables.

P(A) = 0.4, P(B|A) = 0.5, and P(B|¬A) = 0.2 indicate that the probability of B is affected by whether A is true or not. This shows a conditional dependency between A and B.