Prolog Introduction. Gunnar Gotshalks PI-1

Transcription

1 Prolog Introduction PI-1

2 Physical Symbol System Hypothesis A physical symbol system has the necessary and sufficient means for general intelligent action. Allen Newell and Herbert A. Simon PI-2

3 Physical Symbol System Hypothesis A physical symbol system has the necessary and sufficient means for general intelligent action. Allen Newell and Herbert A. Simon Prolog can represent and manipulate symbols PI-3

4 Physical Symbol System Hypothesis A physical symbol system has the necessary and sufficient means for general intelligent action. Allen Newell and Herbert A. Simon Prolog can represent and manipulate symbols Prolog systems can implement general intelligent actions PI-4

5 What is symbolic computing? Computing on» Non-numbers» Non-character-string» Use atoms instead of numbers and strings PI-5

6 What is symbolic computing? 2 Computing on» Non-numbers» Non-character-string» Use atoms instead of numbers and strings Building structures from atoms» Lists, trees, terms, clauses, propositions, etc. PI-6

7 What is symbolic computing? 3 Computing on» Non-numbers» Non-character-string» Use atoms instead of numbers and strings Building structures from atoms» Lists, trees, terms, clauses, propositions, etc. Computing that uses descriptions and creates descriptions PI-7

8 Declarative (Prolog-like) Palindrome example palindrome ( X ) if reverse ( X, X ). X is a palindrome if the reverse of X is X PI-8

9 Declarative (Prolog) Reverse example reverse ( [], [] ). Reverse of empty is empty reverse ( W ^ X, Y ) Reverse of W^X is Y if reverse (X, Z) if Reverse of X is Z and append ( Z, W, Y ). and appending W to Z is Y Reverse of Y W X is reverse of X W Z PI-9

10 Alternatives to symbolic AI Connectionist models» Based on brain metaphor» Model individual neurons and their connections» Properties > Parallel, distributed, sub-symbolic» Examples > Neural nets, associative memories PI-10

11 Alternatives to symbolic AI 2 Emergent models» Based on an evolution metaphor» Potential solutions compete and evolve» Properties > Massively parallel > Complex behaviour evolves out of simple behaviour» Examples > Genetic algorithms > Cellular automata > Artificial life PI-11

12 What is a Prolog Program? A program consists of a database containing one or more facts > A fact is a relationship between a collection of objects» dog ( fido ). > Fido is a dog it is true that Fido is a dog» mother ( mary, joe ). > Mary is the mother of Joe it is true that Mary is the mother of Joe» compete ( ali, leila, tennis). > Ali and Leila compete in tennis it is true that Ali and Leila compete in tennis PI-12

13 What is a Prolog Program? 2 Relationships can have any number of objects Names are usually chosen to be meaningful» Within Prolog, names are just arbitrary strings. It is people who give meaning to names. PI-13

14 What is a Prolog Program? 3 And a program consists of a database of zero or more rules > A rule is an if...then relationship of facts» use ( umbrella ) :- weather ( raining ). > use an umbrella if weather is raining» use ( umbrella ) :- weather ( raining ), own ( umbrella ). > use an umbrella if weather is raining and you own an umbrella» use ( umbrella ) :- weather ( raining ), ( own ( umbrella ) ; borrow ( umbrella ) ). > use an umbrella if weather is raining and you either own an umbrella or can borrow an umbrella PI-14

15 More on rules Rules have the general structure head :- body» Only one fact can be in the head the consequent» The body is a boolean combination of predicates» Use, (and) and ; (or) and () (parenthesis) to logically organize the "condition" the antecedent Rules are written backwards to» emphasize the backward chaining for database search» be more regular in structure, since the head is only one predicate PI-15

16 Constants Constants are names that begin with lower case letters» ali, leila, tennis, dog, fido, mother, mary, joe, umbrella, raining, weather, own, borrow» names of relationships are constants PI-16

17 Variables In place of constants in facts and rules one can have variables» variables are names that begin with upper case letters > X, Y, Who, Whom, List, Person loves ( Everyone, barney ). > Everyone loves barney for all values of Everyone it is the case that loves(everyone, barney) is true. noisy ( Singer ) :- valkyrie ( Singer ) ; tenor ( Singer ). > A Singer is noisy if they are a Valkyrie or a tenor PI-17

18 Variables 2 dwarf ( Person ) :- brother ( Person, Other ), dwarf ( Other ). > A person is a dwarf, if they the brother of other and the other is a dwarf» Variables can also begin with _ (underscore) _ (anonymous variable) _1 _abc (not anonymous variable ) PI-18

19 Running a Prolog Program Programs are stored in one or more files that are consulted On Prism to run SWI Prolog enter % swipl The following prompt appears?- Consult the appropriate file(s) add to the database?- consult ( 'ring.pl ). > SWI-prolog does not have a reconsult predicate, only consult is used > The following is an abbreviation?- [ ring, tower, 'utilities.pl' ]. PI-19

20 Running a Prolog Program 2 Make zero or more queries (next slides) Exit prolog?- CTRL-d /* and for consult ( user ) */ consult ( user ) enables you to enter facts & rules into the database without storing them in a file. It is not an effective way to work with Prolog, as it is error prone PI-20

21 Queries A query in Prolog is boolean combination of predicates like the antecedent of a rule > A query is like a rule, except we leave out the consequent true true :- dwarf ( alberich ). > becomes simply dwarf ( alberich ). Use comma (and), semicolon (or) and parenthesis to form a query expression Most common is to have a single predicate PI-21

22 Queries 2 Answer is a binding of the variables that make the query expression true if no variables then the answer is yes. If no such binding exists, the answer is no The database is searched to match the query. The search» Uses backward chaining» is depth first» is sequential through the database from first to last Try the exercise on ring.pl PI-22

23 Structures Structures are a means of grouping a collection of other objects» Structures are also called compound terms, or complex terms» The name of a structure is called a functor» The items within a structure are called components The general pattern is functor ( component_1, component_2,... component_n ) PI-23

24 Structures 2 Components can also be structures recursive definition If component_1 = functor1 ( comp1, comp2 ) > giving functor ( functor1 ( comp1, comp2 ), component_2,... component_n ) > from functor ( component_1, component_2,... component_n ) PI-24

25 Example structures Books have authors and titles, so we could have book ( dickens, great_expectations ) People have books. In particular, Leila could have Great Expectations has ( leila, book ( dickens, great_expectations ) ) Facts in Prolog are structures where the predicate is the functor of a structure and the arguments of the predicate are the components of the structure PI-25

26 Characters Prolog is based on the ASCII character set Characters are treated as small integers Characters may be» printed» read from a file or keyboard» compared» take part in arithmetic operations Characters are distinguished as» printing visible on the paper» nonprinting look like whitespace PI-26

27 Operators All operators in Prolog are functors, even, ; and :- > A rule such as dwarf ( Person ) :- brother ( Person, Other ), dwarf ( Other ). > is a shorthand for :- ( dwarf ( Person ),, ( brother ( Person, Other ), dwarf ( Other ) ) ). PI-27

28 Operators 2 Arithmetic and relational operators are also functors, thus a + b * c internally is + ( a, * ( b, c) ) This is inconvenient so Prolog permits operators to be written in standard infix notation» You will learn later how you can define your own infix operators PI-28

29 Arithmetic The arithmetic operators do not do arithmetic. No assignments are made > It is simply pattern matching infix operators are simply a convenience for expressing a structure 5 = ==> no = ==> yes = ==> no > Use the operator is to do arithmetic 5 is ==> yes is ==> no Arithmetic is only done on the right! Right hand side is evaluated using arithmetic, then a pattern match is made with the left hand side. PI-29

30 Arithmetic 2 Can use variables in arithmetic expressions for pattern matching A = ==> A has the pattern "4+1" spaces removed A is ==> A has as value the pattern 5 > In some Prologs the latter expression simply responds yes, so try the following. A is 4 + 1, A = 5. ==> A = 5 is the binding for true > More complex example B is 3 + 2, C is B * 5, A is C + B ==> B = 5, C = 25, A =30 PI-30

31 Lists Lists are a ubiquitous structure in many programming languages. The syntax changes (to protect the innocent?)» Actually () are used to delimit structure components and to provide precedence for operators, so using them for lists as well would be confusing. The structure is [ item-1, item-2,..., item-n ] [ a, b, c ] [ a, [ b, c ], [ [ [ d ] ] ], e, [ ] ] The empty list is [ ] PI-31

32 Lists - 2 Lists have a head (first) and a tail (rest), thus [ Head Tail ] [ a, b, c, d ] = [ a [ b, c, d ] ] [ a, b, c, d ] = [ a, b [ c, d ] ] You do not have operators to extract the head and tail, all you have is pattern matching» [ a, b, c, d ] = [ First, Second Rest ] First = a Second = b Rest = [ c, d ]» We will look at example Prolog utilities on lists to demonstrate Empty list has no head or tail [ ] [ _ _ ] PI-32

33 Equal pattern matching» Standard match > A = B > A \= B» Arithmetic values > A is expression > expr1 =:= expr2 equal values expr1 =\= expr2 not equal values PI-33

34 Strict pattern match» == structures identical» \== structures not identical > f(a, b) == f(a, b) > yes > f(a, b) == f(a, X) > no PI-34

35 SWIPL Help You can get help with the following predicates. help. Brings up a help window from which you can search for information help(functor). Brings up a help window for the predicate with the name functor help(is). Brings up a help window for the predicate named is. PI-35