Artificial Intelligence Reasoning and Knowledge Systems

Knowledge Representation: Frames and Semantic Networks

Frames

Frame: Represents knowledge using frames; it is more structured and object-oriented (OO). It is better for representing detailed knowledge about entities and describes an object or situation in detail. Each frame contains slots (attributes) and fillers (values). The basic unit is the frame.

Disadvantages:

• Limited flexibility
• Inheritance problems
• Difficulty representing complex relationships

Semantic Networks

Semantic Network: Represents knowledge as a graph of concepts connected by relationships. It is more graphical and relationship-oriented, making it better for representing associations between concepts. It shows relationships among concepts and consists of nodes and edges. The basic unit consists of nodes and links.

Disadvantages:

• Ambiguity
• Maintenance challenges
• Difficulty representing complex knowledge

Reasoning: Forward and Backward Chaining

Forward Chaining

Forward Chaining starts from known facts; it is data-driven. It generates all possible conclusions and is better for monitoring systems. It is best when all facts are available and every possible conclusion is required.

Backward Chaining

Backward Chaining starts from a goal; it is goal-driven. It searches only for the required conclusion and is better for expert systems. It is best when a specific goal or query needs to be verified efficiently.

Search Strategies: Uninformed and Informed

Uninformed Search

Uninformed Search uses no heuristic information and explores nodes blindly. It may search unnecessary nodes and is slower for large search spaces. Examples include Breadth-First Search (BFS) and Depth-First Search (DFS).

Informed Search

Informed Search uses heuristic information and explores promising nodes first. It usually searches fewer nodes and is more efficient and faster. An example is A* search.

AI Approaches: Thinking Humanly vs. Rationally

Thinking Humanly

Thinking Humanly: Mimics the way humans think; it may include human errors and biases. It focuses on understanding human intelligence and is based on cognitive science and human thought processes. An example is a cognitive simulation system.

Thinking Rationally

Thinking Rationally: Uses logical reasoning to make decisions. It always aims for the most logical solution. It focuses on providing correct conclusions and is based on the laws of logic and rational processing.

Reasoning Methods: Deductive and Inductive

Deductive Reasoning

Deductive Reasoning starts with general facts or rules. It is top-down reasoning used in expert systems. It produces certain conclusions if the premises are true.

Inductive Reasoning

Inductive Reasoning starts from specific observations. It is bottom-up reasoning used in Machine Learning (ML). It produces probable conclusions.