Lernzettel: Artificial Intelligence Question Bank

📋 Course Outline

1. Intelligent Agents and Environments
2. Search-Based Problem Solving
3. Heuristics and Online Search
4. Game Playing and CSP
5. Logical Reasoning and Knowledge Agents
6. Probabilistic Reasoning

📖 1. Intelligent Agents and Environments

🔑 Key Concepts & Definitions

• Intelligent agent : Perceives its environment through sensors and acts on that environment through actuators.
• PEAS framework : Specifies an agent's Performance measure, Environment, Actuators, and Sensors.
• Rational agent : Selects actions expected to maximize its performance measure given its percept sequence and available knowledge.
• Agent architecture : Intelligent agent architectures determine how percepts are transformed into actions, and they include simple reflex, model-based, goal-based, and utility-based designs.

📝 Essential Points

• Task environments may be deterministic, where the next state is fixed by the current state and action, or stochastic, where outcomes involve uncertainty.

📖 2. Search-Based Problem Solving

★ Must-know

📌 Uninformed search uses no problem-specific heuristic information, whereas informed search uses additional knowledge to guide exploration.

• Breadth-First Search expands the shallowest unexpanded nodes before deeper nodes in the search tree.

• Depth-First Search expands the deepest unexpanded node and backtracks when a branch cannot produce a solution.

• Uniform Cost Search expands the frontier node with the lowest accumulated path cost.

Further detail

• A problem-solving agent formulates a goal and state-space problem, then searches for an action sequence that reaches a goal state.

📖 3. Heuristics and Online Search

🔑 Key Concepts & Definitions

• Heuristic function : Estimates the cost of reaching a goal from a given state and guides heuristic search.

★ Must-know

• A* search combines the cost accumulated to reach a state with a heuristic estimate of the remaining cost to prioritize nodes.

• Hill climbing repeatedly moves from the current state to a better neighboring state in an optimization problem.

Further detail

• Simulated annealing is a local search method that can accept some worsening moves while gradually reducing that tolerance.

• An online search agent interleaves acting and searching while discovering an initially unknown environment.

📖 4. Game Playing and CSP

🔑 Key Concepts & Definitions

• Constraint satisfaction problem : Consists of variables, their possible domains, and constraints restricting allowed combinations of values.

★ Must-know

• The minimax strategy selects a move that maximizes the best outcome against an opponent assumed to minimize the agent's outcome.

• Alpha-beta pruning removes game-tree branches that cannot affect the final minimax decision.

• Backtracking search assigns CSP variables incrementally and returns to an earlier assignment when a constraint is violated.

Further detail

• Constraint propagation removes domain values that cannot participate in a consistent CSP assignment.

💡 Memory Hook

Games optimize decisions; CSPs satisfy constraints

📖 5. Logical Reasoning and Knowledge Agents

🔑 Key Concepts & Definitions

• Knowledge-based agent : Stores representations of facts and rules and uses inference to choose actions.
• Propositional logic : Represents knowledge as propositions that are either true or false and combines them with logical connectives.
• First-Order Logic : Represents objects, predicates, relations, functions, and quantified statements about a domain.

★ Must-know

• Forward chaining repeatedly applies rules whose premises match known facts to derive new conclusions.

Further detail

• Backward chaining starts with a goal and recursively searches for rules and facts that could establish it.

📖 6. Probabilistic Reasoning

🔑 Key Concepts & Definitions

• Uncertainty : In artificial intelligence occurs when an agent cannot determine the true state or outcome with complete confidence.
• Bayesian inference : Updates the probability of hypotheses using evidence and prior beliefs.
• Bayesian network : Represents probabilistic dependencies among variables using a directed graphical structure.

📝 Essential Points

📌 Exact inference calculates probabilities through an exact computation, whereas approximate inference estimates them using methods suitable for large Bayesian networks.

• Naive Bayes classification applies Bayesian reasoning to assign a class using observed features under a conditional-independence assumption.

📊 Synthesis Tables

Search Strategy Comparison

Strategy	Guidance	Selection principle
Breadth-First Search	Uninformed	Shallowest node
Depth-First Search	Uninformed	Deepest node
Uniform Cost Search	Uninformed	Lowest path cost
A* Search	Informed	Path cost plus heuristic

Inference Method Comparison

Method	Starting point	Main operation
Forward chaining	Known facts	Derive consequences
Backward chaining	Goal	Find supporting rules and facts
Exact inference	Probability model	Compute probabilities exactly
Approximate inference	Probability model	Estimate probabilities