This preview shows page 1. Sign up to view the full content.
Unformatted text preview: needed.
Terms of regular use
/ Syntax of a representation specifies the symbols that may be used and the ways that
they may be arranged.
/ Semantics of a representation specifies how meaning is embodied in the symbols and
the symbol arrangement allowed by the syntax.
/ A representation is a set of syntactic and semantic conventions that make it possible
to describe things.
The main knowledge representation techniques used in AI are formal grammars and languages,
predicate logic, production rules, semantic nets, and frames. Features and descriptions are
usually not considered knowledge representations, added for practical reasons.
Cannot be considered pure knowledge but are nevertheless used for representing knowledge as a
part of a more complex representation structure. Descriptions usually represent some scalar
properties of objects, and are called features. Typically a single description is insufficient for
object representation, therefore the description are combined in feature vectors. Numerical
feature vectors are inputs for statistical pattern recognition techniques.
The size feature can be used to represent an area property, and the compactness feature describes
circularity. Then the feature vector x = (size, compactness) can be used for object classification
into the following classes of objects: small, large, circular, noncircular, small and circular, small
and noncircular,, etc… information about small/large and circular/noncircular must be available.
If an object’s structure needs to be described, feature description is not appropriate. A structural
description is formed from existing primitives and the relations between them.
One object can be described by a chain, a tree, a graph, etc.. of symbols. Nevertheless, the whole
class of objects cannot be described by single chain, a single tree etc., but a class of structurally
described objects can be represented by grammars and languages. Grammars and languages
(similar to natural languages) provide rules defining how the chains, trees, or graphs can be
constructed from a set of symbols (primitives) Primitives are represented by information about
Predicate logic plays a very important role in knowledge representation- it introduces a
mathematical formalism to derive new knowledge by applying a mathematical deduction.
Predicate logic works with combinations of logic variables, quantifiers (∃,∀), and logic operators
(and, or ,not, implies, equivalent). The logic variables are binary (true, false). The idea of proof
and rules of inference such as modus ponens and resolution are the main building block of
Predicate logic forms the essence of the programming language PROLOG, which is widely used
if objects are described by logic variables. Requirements of “pure truth” represent the main
weakness of predicate logic in knowledge representation, since it does not allow work with
uncertain or incomplete information.
Faculty of Engineering Robotics Technology MECH 4041 B. Eng (Hons.) Mechatronics S. Venkannah Mechanical and Production Engineering Department Production rules
Production rules represent a wide variety of knowledge representations that are based on
condition ratio pairs. The essential model of behavior of a system based on production rules ( a
production system) can be described as follows:
If condition X holds then action Y is appropriate
Information about what action is appropriate at what time represents knowledge. The procedural
character of knowledge represented by production rules is another important property- not all the
information about the objects must be listed as an object property. Consider a simple knowledge
base where the following knowledge is present:
If ball then circular
Let the knowledge base also include the statements
Object A is-a ball
Object B is-a ball
Object C is-a shoe
To answer the question how many balls are circular?, if enumerative knowledge is used, the
knowledge must be listed as
Object A is-a (ball, circular)
Object B is-a (ball, circular)
If procedural knowledge is used, the knowledge base together with he knowledge gives the same
information in a significantly more efficient manner.
Both production rule knowledge representation and production systems appear frequently in
computer vision and image understanding problems.
Fuzzy logic has been developed to overcome the obvious limitations of numerical or crisp
representation of information. Consider the use of knowledge represented by equation
If ball then circular
For recognition of balls; using the production rule, the knowledge about balls may be represented
If circular then ball
If the object in a two dimensional image is considered circular then it may represent a ball. But
balls may not be perfectly circular, thus it is necessary to define some circularity threshold so
that all reasonably circular objects from our set of objects are labeled as balls
How circular must a ball be to b...
View Full Document
This document was uploaded on 03/12/2014 for the course MECHANICAL 214 at University of Manchester.
- Spring '14