{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

AI Spring 2010 Lecture 8-9

AI Spring 2010 Lecture 8-9 - Artificial Intelligence...

Info iconThis preview shows pages 1–9. Sign up to view the full content.

View Full Document Right Arrow Icon
Click to edit Master subtitle style Artificial Intelligence Lectures 8&9: Logical Agents (Part III: Knowledge & Reasoning, Chapter 7) Spring 2010 Instructor: Paul S. Rosenbloom
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
22 Project 1 £ Available on Blackboard £ Due by end of day (11:59pm PST) on 2/23 l Penalty for late submission (after late days used) £ Solving Sudoko l Using greedy best-first search and steepest-ascent hill climbing l Two different heuristic functions l Some short questions for you to answer as well £ TA (Hyokyeong Lee) should be first line for questions about programming and can answer content-oriented questions as well l I can answer content-oriented questions as well
Background image of page 2
33 This Week’s Lectures £ Knowledge-based agents £ Wumpus world £ Logic in general - models and entailment £ Propositional (Boolean) logic l°°° £ Equivalence, validity, satisfiability £ Inference rules and theorem proving Quite a bit of new terminology this week
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
44 Knowledge-Based Agents £ So far we have seen agents use several problem-solving-oriented kinds of knowledge l Problem spaces (states and actions, models of E) l Including sensory information about states l Problems (initial and goal states) l Search algorithms (K about how to search) l Evaluation, utility, and objective functions l Control rules and islands (brief discussion) £ The focus has been on knowledge encoded (or formulated) for a particular manner of use l More procedurally encoded
Background image of page 4
55 Knowledge-Based Agents (2) £ Starting now we focus on knowledge more broadly l How is it represented for a broad set of possible uses? l Not just to for a single class of problems l But for a range problems, communicating, reading, teaching, etc. Without the inherent ambiguity in natural language “John saw a diamond through the window and coveted it.” “John threw a brick through the window and broke it.” l Support agent generality r Syntax specifies the form of the knowledge l Semantics determines the meaning of the knowledge l How do we reason with it to answer general questions? l Combine multiple facts to derive new ones that are implicit All men are mortal AND Socrates is a man ° Socrates is mortal l Inference processes syntax so as to reflect/respect semantics
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
66 Knowledge-Based Agents (3) £ Inside the box (replacing the “?”) is: l A knowledge base (KB) of content l Both general and domain specific information l In its simplest form can think of it as enhanced state content l An inference engine (or knowledge architecture) l A domain-independent means of deriving conclusions from KB £ Need to understand both how to build inference engines and how to represent knowledge
Background image of page 6
77 Knowledge Bases £ Knowledge base = set of sentences in a formal language £ I.e., in a knowledge representation language with a well defined syntax and semantics l £ Declarative approach to building agents (or other systems) l Tell KB what it needs to know l Ask KB for answers to questions Knowledge Base Tell Ask
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
88 A Simple Knowledge-Based Agent £ The agent must be able to: l Represent states, actions, etc.
Background image of page 8
Image of page 9
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}