© Franz J. Kurfess ES Implementation 1 CPE/CSC 481: Knowledge-Based Systems Dr. Franz J. Kurfess Computer Science Department Cal Poly
© Franz J. Kurfess ES Implementation 2 Usage of the Slides u these slides are intended for the students of my CPE/CSC 481 “Knowledge-Based Systems” class at Cal Poly SLO u if you want to use them outside of my class, please let me know u I usually put together a subset for each quarter as a “Custom Show” u to view these, go to “Slide Show => Custom Shows”, select the respective quarter, and click on “Show” u To print them, I suggest to use the “Handout” option u 4, 6, or 9 per page works fine u Black & White should be fine; there are few diagrams where color is important
© Franz J. Kurfess ES Implementation 3 Course Overview u Introduction u Knowledge Representation u Semantic Nets, Frames, Logic u Reasoning and Inference u Predicate Logic, Inference Methods, Resolution u Reasoning with Uncertainty u Probability, Bayesian Decision Making u Expert System Design u ES Life Cycle u CLIPS Overview u Concepts, Notation, Usage u Pattern Matching u Variables, Functions, Expressions, Constraints u Expert System Implementation u Salience, Rete Algorithm u Expert System Examples u Conclusions and Outlook
© Franz J. Kurfess ES Implementation 4 Overview Implementation of Rule-Based Systems u Motivation u Objectives u Chapter Introduction u Important Concepts u Performance Aspects u Pattern Matching u Basic Idea u Unification u Pattern Matching in Rule- Based Systems u Rete Algorithm u Overview u Rete Network u Assert and Retract u Optimizations u Improvements u Rule Formulation u General vs. Specific Rules u Simple vs. Complex Rules u Loading and Saving Facts u Important Concepts and Terms u Chapter Summary
© Franz J. Kurfess ES Implementation 5 Logistics u Introductions u Course Materials u textbooks (see below) u lecture notes u PowerPoint Slides will be available on my Web page u handouts u Web page u u Term Project u Lab and Homework Assignments u Exams u Grading
© Franz J. Kurfess ES Implementation 6 Bridge-In
© Franz J. Kurfess ES Implementation 7 Pre-Test
© Franz J. Kurfess ES Implementation 8 Motivation u pattern matching and unification are powerful operations to determine the similarity and consistency of complex structures u they are at the core of many rule-based and predicate logic mechanisms u their application goes beyond rule-based systems u study concepts and methods that are critical for the functionality and performance of rule-based systems u pattern matching and the Rete algorithm u use and formulation of rules
© Franz J. Kurfess ES Implementation 9 Objectives u comprehend the mechanics of pattern matching in rule-based systems u basic concepts and techniques u Rete algorithm u understand the effects of matching and rule formulation on the performance of rule-based systems u learn to write rule-based programs and implement systems in an efficient way u analyze and evaluate the performance of rule-based programs and systems u identify bottlenecks u formulate and implement strategies for performance improvements
© Franz J. Kurfess ES Implementation 10 Evaluation Criteria
© Franz J. Kurfess ES Implementation 11 Overview Implementation of Rule-Based Systems u due to their more declarative nature, it can be difficult to evaluate and predict the performance of rule-based systems u time to complete a task u memory usage u disk space usage u pattern matching can be used to eliminate unsuitable rules and facts u but it can also introduce substantial overhead
© Franz J. Kurfess ES Implementation 12 Chapter Introduction u Important Concepts u entities with internal structure u data structures, objects, components u terms, sentences, graphs u diagrams, images u concepts, hierarchies u Performance Aspects u somewhat different from conventional programs u less control over the runtime behavior u pattern matching can do a lot of the work
© Franz J. Kurfess ES Implementation 13 Pattern Matching u determines if two or more complex entities (patterns) are compatible with each other u patterns can be (almost) anything that has a structure u pictures: mugshot vs. person u drawings: diagrams of systems u expressions: words,sentences of a language, strings u graphs are often used as the underlying representation u the structure of the graphs must be compatible u usually either identical, or one is a sub-graph of the other u the individual parts must be compatible u nodes must have identical or compatible values uvariables are very valuable u links must indicate compatible relationships u compatibility may be dependent on the domain or task [Giarratano & Riley 1998, Friedmann-Hill 2003, Gonzalez & Dankel, 2004]
© Franz J. Kurfess ES Implementation 14 ????? Pattern Matching Example u images u Do both images refer to the same individual? u Do they have other commonalities? v test Bucky Bucky likes fish Bucky Bucky likes fish Bucky and Satchel Satchel likesBucky
© Franz J. Kurfess ES Implementation 15 Pattern Matching Example u shapes ?? ?????
© Franz J. Kurfess ES Implementation 16 Pattern Matching Examples u constants and variables “Hans” “Franz” ? “Josef” “Joseph” first_name “Joseph” ? last_name “Joseph”
© Franz J. Kurfess ES Implementation 17 Pattern Matching Examples u terms u composed of constants, variables, functions ? father(X) “Joseph” ? father(Y)father(X) mother(X) ?? father(father(X))grandfather(X)
© Franz J. Kurfess ES Implementation 18 Unification u formal specification for finding substitutions that make logical expressions identical u the unification algorithm takes two sentences and returns a unifier for them (if one exists) Unify(p,q) = if Subst( ,p) = Subst( ,q) u if there is more than one such substitution, the most general unifier is returned u used in logic programming, automated theorem proving u possibly complex operation u quadratic in the size of the expressions u “occur check” sometimes omitted v determines if a variable is contained in the term against which it is unified
© Franz J. Kurfess ES Implementation 19 Pattern Matching in Rule-Based Systems u used to match rules with appropriate facts in working memory u rules for which facts can be found are satisfied u the combination of a rule with the facts that satisfy it is used to form activation records v one of the activation records is selected for execution
© Franz J. Kurfess ES Implementation 20 Simplistic Rule-Based Pattern Matching u go through the list of rules, and check the antecedent (LHS) of each rule against the facts in working memory u create an activation record for each rule with a matching set of facts u repeat after each rule firing u very inefficient u roughly (number of rules) * (number of facts) (number of patterns) u the actual performance depends on the formulation of the rules and the contents of the working memory
© Franz J. Kurfess ES Implementation 21 Rete Algorithm u in most cases, the set of rules in a rule-based system is relatively constant u the facts (contents of working memory) change frequently u most of the contents of working memory, however, don’t change every time u optimization of the matching algorithm u remember previous results u change only those matches that rely on facts that changed u the Rete algorithm performs an improved matching of rules and facts u invented by Charles Forgy in the early 80s u basis for many rule-based expert system shells [ Friedmann-Hill 2003, Giarratano & Riley 1998, Friedmann-Hill 2003, Gonzalez & Dankel, 2004]
© Franz J. Kurfess ES Implementation 22 Rete Network u the name comes from the latin word rete u stands for net u consists of a network of interconnected nodes u each node represents one or more tests on the LHS of a rule u input nodes are at the top, output nodes at the bottom u pattern nodes have one input, and check the names of facts u join nodes have two inputs, and combine facts u terminal node at the bottom of the network represent individual rules u a rule is satisfied if there is a combination of facts that passes all the test nodes from the top to the output node at the bottom that represents the rule u the Rete network effectively is the working memory for a rule- based system
© Franz J. Kurfess ES Implementation 23 Rete Network Example 1 (deftemplate x (slot a)) (deftemplate y (slot b)) (defrule example-1 (x (a ?v1)) (y (b ?v1)) ==> ) ?= x ?= y Left.0.a ?= Right.b example-1 ?v1 ?v1 = ?v1
© Franz J. Kurfess ES Implementation 24 Rete Left and Right Memories u left (alpha) memory u contains the left input of a join node u right (beta) memory u contains the right input of a join node u notation: Left.p.q ?= Right.r u compare the contents of slot q in fact p from the left memory with slot r in the fact from the right memory (deftemplate x (slot a)) (deftemplate y (slot b)) (defrule example-1 (x (a ?v1)) (y (b ?v1)) ==> ) ?= x ?= y Left.0.a ?= Right.b example-1 ?v1 ?v1 = ?v1
© Franz J. Kurfess ES Implementation 25 Running the Network only facts x and y are considered all facts where x.a == y.b pass the join network all {x, y} tuples are fowarded to the next node u compare the contents of slot q in fact p from the left memory with slot r in the fact from the right memory (deftemplate x (slot a)) (deftemplate y (slot b)) (defrule example-1 (x (a ?v1)) (y (b ?v1)) ==> ) ?= x ?= y Left.0.a ?= Right.b example-1 ?v1 ?v1 = ?v1
© Franz J. Kurfess ES Implementation 26 Rete Network Example 2 u shares some facts with Example 1 (deftemplate x (slot a)) (deftemplate y (slot b)) (deftemplate z (slot c)) (defrule example-2 (x (a ?v2)) (y (b ?v2)) (z) ==> ) ?= x ?= y ?= z Left.0.a ?= Right.b example-2 ?v2 ?v2 = ?v2 ?v2
© Franz J. Kurfess ES Implementation 27 Rete Network Example 2 with Assert u additional fact asserted (deftemplate x (slot a)) (deftemplate y (slot b)) (deftemplate z (slot c)) (defrule example-2 (x (a ?v2)) (y (b ?v2)) (z) ==> ) (assert (z (c 17)) ?= x ?= y ?= z Left.0.a ?= Right.b example-2 ?v2 ?v2 = ?v2 ?v2 ?v2 = 17 17
© Franz J. Kurfess ES Implementation 28 Assert and Retract with Rete u asserting additional facts imposes some more constraints on the network u retracting facts indicates that some previously computed activation records are not valid anymore, and should be discarded u in addition to the actual facts, tags are sent through the networks ADD to add facts (i.e. for assert) REMOVE to remove facts (i.e. for retract) CLEAR to flush the network memories (i.e. for reset) UPDATE to populate the join nodes of newly added rules v already existing join nodes can neglect these tokens
© Franz J. Kurfess ES Implementation 29 Rete Network Optimization u networks with shared facts can be combined (deftemplate x (slot a)) (deftemplate y (slot b)) (deftemplate z (slot c)) (defrule example-1 (x (a ?v1)) (y (b ?v1)) ==> ) (defrule example-2 (x (a ?v2)) (y (b ?v2)) (z) ==> ) ?= x ?= y ?= z Left.0.a ?= Right.b example-1example-2
© Franz J. Kurfess ES Implementation 30 Further Optimizations u sophisticated data structures to optimize the network u hash table to presort the tokens before running the join node tests u fine-tuning via parameters u frequently trade-off between memory usage and time
© Franz J. Kurfess ES Implementation 31 Special Cases for Pattern Matching u additional enhancements of the Rete network can be used to implement specific methods u backward chaining v requires a signal indicating to the network that a particular fact is needed not conditional element v indicates the absence of a fact v requires special join nodes and special fields in the tokens passing through the network test conditional element v uses a special join node that ignores its right input v the result of the function is passed on
© Franz J. Kurfess ES Implementation 32 Exploring the Rete Network in Jess (watch compilations) function diagnostic output when rules are compiled example-1: t +1 one-input (pattern) node added to the Rete network +2 two-input (pattern) node added +t terminal node added (view) function u graphical viewer of the Rete network in Jess (matches ) function u displays the contents of the left and right memories of the join nodes for a rule u useful for examining unexpected rule behavior
© Franz J. Kurfess ES Implementation 33 Rule Formulation u Pattern Order u General vs. Specific Rules u Simple vs. Complex Rules u Loading and Saving Facts [Giarratano & Riley 1998]
© Franz J. Kurfess ES Implementation 34 Pattern Order u since Rete saves information about rules and facts, it can be critical to order patterns in the right way u otherwise a potentially huge number of partial matches can be generated
© Franz J. Kurfess ES Implementation 35 Example Pattern Order (deffacts information (find-match a c e g)f1 (item a)f2 (item b)f3 (item c)f4 (item d)f5 (item e)f6 (item f)f7 (item g))f8 (defrule match-1 (find-match ?x ?y ?z ?w)P1 (item ?x)P2 (item ?y)P3 (item ?z)P4 (item ?w)P5 ==> (assert (found-match ?x ?y ?z ?w)) (deffacts information (find-match a c e g) (item a) (item b) (item c) (item d) (item e) (item f) (item g)) (defrule match-2 (item ?x) (item ?y) (item ?z) (item ?w) (find-match ?x ?y ?z ?w) ==> (assert (found-match ?x ?y ?z ?w)) [Giarratano & Riley 1998]
© Franz J. Kurfess ES Implementation 36 Pattern Matches full matches P1: f1 P2: f2,f3,f4,f5,f6,f7,f8 P3: f2,f3,f4,f5,f6,f7,f8 P4: f2,f3,f4,f5,f6,f7,f8 P5: f2,f3,f4,f5,f6,f7,f8 u partial matches P1: [f1] P1-2: [f1,f2] P1-3: [f1,f2,f4] P1-4: [f1,f2,f4,f6] P1-5: [f1,f2,f4,f6,f8] Total: 29 full, 5 partial matches u full matches P1: f2,f3,f4,f5,f6,f7,f8 P2: f2,f3,f4,f5,f6,f7,f8 P3: f2,f3,f4,f5,f6,f7,f8 P4: f2,f3,f4,f5,f6,f7,f8 P5: f1 u partial matches P1: [f2,f3,f4,f5,f6,f7,f8] P1-2:[f2,f2],[f2,f3],[f2,f4],[f2,f5], [f2,f6],[f2,f7],[f2,f8], [f3,f2],[f3,f3],[f3,f4],[f3,f5], [f3,f6],[f3,f7],[f3,f8],... P1-3, P1-4:... P1-5: [f2,f4,f6,f8, f1] Total: 29 full, 2801 partial matches
© Franz J. Kurfess ES Implementation 37 Adding another Fact what is the effect on the two cases if another fact (item h) is added? no significant changes for match-1 u in particular, no additional partial matches major changes for match-2 u another 1880 partial matches
© Franz J. Kurfess ES Implementation 38 Guidelines for Pattern Matches u try to formulate your rule such that the number of matches is low u full and partial matches u try to limit the number of old partial matches u removing those also is time-consuming u in general, the state of the system should be reasonably stable u matches u assertion, retraction, modification of facts
© Franz J. Kurfess ES Implementation 39 Guidelines for Pattern Ordering u most specific patterns first u smallest number of matching facts u largest number of variable bindings to constrain other facts u patterns matching volatile facts go last u facts that are changing frequently should be used by patterns late in the LHS u smallest number of changes in partial matches u may cause a dilemma with the above guideline u patterns matching the fewest facts first u reduces the number of partial matches
© Franz J. Kurfess ES Implementation 40 Multifield Variables u multifield wildcards and multifield variables are very powerful, but possible very inefficient u should only be used when needed u limit their number in a single slot of a pattern
© Franz J. Kurfess ES Implementation 41 Test Conditional Element the test conditional element should be placed as close to the top of the rule as possible u reduces the number of partial matches u evaluation of expressions during pattern matching is usually more efficient
© Franz J. Kurfess ES Implementation 42 Built-In Pattern Matching Constraints u the built-in constraints are always more efficient than the equivalent expression u not so good: (defrule primary-color color ?x&: (or (eq ?x red) (eq ?x green) (eq ?x blue) ==> (assert (primary-color ?x))) u better: (defrule primary-color color ?x&red|green|blue) ==> (assert (primary-color ?x)))
© Franz J. Kurfess ES Implementation 43 General vs. Specific Rules u some knowledge can be expressed through many specific, or a few general rules u specific rules generate a top-heavy Rete network with many pattern nodes and fewer join nodes u general rules offer better opportunities for sharing pattern and join nodes u it usually is easier to write an inefficient general rule than an inefficient specific rule
© Franz J. Kurfess ES Implementation 44 Simple vs. Complex Rules u simple rules are sometimes elegant, but not necessarily efficient u storing temporary facts can be very helpful v especially in recursive or repetitive programs
© Franz J. Kurfess ES Implementation 45 Loading and Saving Facts u facts can be kept in a file, and loaded into memory when needed (load-facts) and (save-facts) functions u may lead to visibility or scoping problems if the respective deftemplates are not contained in the current module
© Franz J. Kurfess ES Implementation 46 Figure Example
© Franz J. Kurfess ES Implementation 47 Post-Test
© Franz J. Kurfess ES Implementation 48 Evaluation u Criteria
© Franz J. Kurfess ES Implementation 49 Use of References u [Giarratano & Riley 1998] [Giarratano & Riley 1998] u [Russell & Norvig 1995] [Russell & Norvig 1995] u [Jackson 1999] [Jackson 1999] u [Durkin 1994] [Durkin 1994] [Giarratano & Riley 1998]
© Franz J. Kurfess ES Implementation 50 Important Concepts and Terms u agenda assert u backward chaining u constant u fact u expert system (ES) u expert system shell u forward chaining u join node u knowledge base u knowledge-based system u left (alpha) memory matches u matching u pattern u pattern matching u pattern node u RETE algorithm retract u right (beta) memory u rule u substitution u term test conditional element u unification u variable view u working memory
© Franz J. Kurfess ES Implementation 51 Summary ES Implementation u for rule-based systems, an efficient method for pattern matching between the rule antecedents and suitable facts is very critical u matching every rule against all possible facts repeatedly is very inefficient u the Rete algorithm is used in many expert system shells u it constructs a network from the facts and rules in the knowledge base u since certain aspects of the knowledge base are quite static, repeated matching operations can be avoided u a few strategies can be used by programmers to achieve better performance u most specific patterns first, patterns with volatile facts last u careful use of multifield variables, general rules use of the test conditional element, built-in pattern constraints u loading and saving of facts
© Franz J. Kurfess ES Implementation 52