Home

OPS-2000 Version 2.1 - User`s Manual

1. backward chaining forward chaining normal E F AA A T TE threshold normal confidence factor fuzzy mn mx si sd www siliconvalleyone com Page 23 of 56 OPS 2000 User s Guide Forward Chaining A forward chaining rule has the generic syntax given below defrule lt name gt lt LHS gt gt lt RHS gt A rule s LHS consists of pattern logic When a rule s LHS is satisfied matched an activation of the rule is placed into the agenda The conflict resolution algorithm decides which agenda entry is to be fired When a rule is fired its RHS is executed This can cause assertions and retractions on the knowledge base For example RuleOne defrule RuleOn The monkey is at position 22 gt assert Move monkey to position 23 This is a simple rule but it shows the declarative nature of OPS 2000 rules if the monkey is at position 22 then issue a command to move the monkey to position 23 Of course this is just ou
2. Instance ly ll 4 4 The cycle given above reflects the typical class instance s object life cycle A class object comes from three primary sources it can be a static object in which case it is created by the system when the program begins and destroyed when the program terminates it can be an auto object in which case it is created when a compound statement block is entered and destroyed when the block is exited or lastly it can be a heap object in which case it is created by the new operator and destroyed by the delete operator In the above diagram there are two boxes that correspond to the new and delete operators these boxes respectively serve as the starting and finishing points of an object s life The other two sources of class objects have starting and finishing points that can be respectively swapped for the new operator box and the delete operator box The use of constructors and destructors provides a way to ensure that all class objects are properly initialized and deinitialized Improperly initialized objects tends to be a common source of software errors so OPS 2000 has made it a requirement that each class have an explicitly declared constructor and destructor which is not a current C requirement www siliconvalleyone com Page 10 of 56 OPS 2000 User s Guide Expert Systems Expert systems are computer programs that draw upon the organized expertise of one or more human e
3. WEST UP DOWN DOWN UP Dungeon PNG s Office Each location understands the message move and responds to this message by returning a reference to the location that corresponds to the direction of the specified move For example in the diagram given above location Computer Laboratory only understands the move message UP and returns a reference to location English Building in response to this message and returns an empty reference location to any other move message Each location can contain people and things Each person has a fixed starting location while a dozen things are randomly distributed in the ten locations These dozen things are six pizzas five beers and a magical wand www siliconvalleyone com Page 54 of 56 OPS 2000 User s Guide OPS 2000 Help Desk Defrelation Defrelation definitions are global coming into effect as soon as they have been successfully compiled Therefore any patterns that had been previously compiled as free form which matches a new relation s generic form will still be treated as free form even though they are now of a relation form This is due to the fact that the pattern was compiled as free form and all data objects that would have matched it before the defrelation will now be treated as relation data and pattern matched to a relation s patterns Many unrelated expert ob
4. When a dribble file is open the system writes all system information to the file instead of the standard output video monitor System information includes all output from OPS 2000 specific functions The Microsoft Windows version of open_dribble echos all data sent to the STDIO window to the dribble file Consequently STDIO window data is simultaneously displayed on the window and written to the dribble file This acts similar to the script command found in some operating systems A dribble example is given below gt gt load dilemma ops load farmer s dilemma gt watch rules watch rules gt open dribble one dbl 0 open dribble file truncate gt reset reset the system gt run 1 run the system until it stops gt close dribble close the dribble file gt In this example all rules watch information is written to the dribble file A dribble file must be closed to ensure that all dribble information has been written to the file When using a nonWindows version of OPS 2000 this will create a file with a trace of the problem s rule firings The rule firings are not displayed on the screen If the problem had used the print function for standard output instead of the printf function then the problem s output would also appear in the dribble file and not on the screen www siliconvalleyone com Page 40 of 56 OPS 2000 User s Guide Function Calls One of OPS 2000
5. and a list of moves The list of moves has a condition for each movement the list is checked sequentially until a move occurs www siliconvalleyone com Page 52 of 56 OPS 2000 User s Guide student Students live in their favorite working environments When a student moves they will either pick up something at the current location or move to another location adjacent to their current location 1 If there are things at the current location pick one up 2 Otherwise randomly move from the current location dean The dean lives in the dean s office and when restless searches the campus for beers When the dean smashes a single beer the dean returns to the dean s office The dean first checks a location for beers and if there are none performs a beer check on a random person in the room 1 If there are any beers at this location pick one up smash it and return to the dean s office 2 If there are any people at this location then randomly select one and perform a beer check If the person has beer then take one smash it and return to the dean s office 3 Otherwise randomly move from the current location troll The troll lives in the dungeon and eats pizzas and people The troll prefers a good pizza over a person 1 If there are any pizzas at this location then pick up one eat it and return to the dungeon for some rest 2 Otherwise if there are any people at this location randomly select one to eat If the person h
6. purpose are limited to a period of thirty days from date of purchase Silicon Valley One shall not be liable for indirect special or consequential damages resulting from the use of this product Some states do not allow the exclusion or limitation of incidental or consequential damages or the above limitations might not apply to you This warranty gives you specific legal rights and you may also have other rights which vary from state to state www siliconvalleyone com Page 2 of 56 OPS 2000 User s Guide Table of Contents INtFoOdU CON armarse 5 OVERVIEW DEPA OPEENS EN E TEE MAE gisele EA SAE PARO a dl e lle lio de ciu para lll ines E 5 INNOVA TON e e e e A a ae ee col aches 5 GREATION EAE AD AE TEPER EAE ENANA cereales cena ab eii Lea add aaa oro ia 5 NOTES Lt A A aia 6 Mantal Nota lO A A n 7 Tutorial Getting Stadium 8 Object Oriented Programming uc 10 AC r CASS INSTANCE S UE CYCLE additional 10 Expert Systems iiiar ariassan ariaa aai araa araara ai aau arias ipae asiaan 11 Prod cti on Systems id da 12 BYNEI N EAA E EAA e a e AAE A E e e ta tl led EAE 12 GOA DIRECTE Dix ta A A E A a e a a a 12 COMPONENTS OF A PRODUCTION SYSTEM 0 0 0s0secesesseeeeeeeeeeeseeeeecesseeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeseeeseeeeeeeeeeeeeeeeeeeeees 13 Productions 221 A a a o ad Sai dre ed th Als ene ido iria e CAS 13 WORO MEON a o ecelies 13 ernro tor rensei eet ee ahi eee thee eh ele bide ie Hee Heath ee thei 13 OPS 2000 Pre Eur 14 CO
7. LHS goal pattern which is matched against the GWM and a RHS which is pattern logic This rule says that if Rudolph has already agreed to give his services as a headlight then a matching LHS goal is proven www siliconvalleyone com Page 48 of 56 OPS 2000 User s Guide The Reindeer Keeper anxiously waiting Santa s orders to lift off reads the goal asking whether Rudolph will give his services as a headlight for Santa s sleigh In a flash of a second he runs over to Rudolph and asks the big question LE ReindeerKeeper if defrule ReindeerKeeper fc match goal priority 99 Below that of rule RudolphsNose Sgoal lt goal rudolphs nose symbol answer query Rudolph will you guide Santa s sleigh tonight gt amp answer if answer yes answer y retract goal else refute goal ReindeerKeeper The above forward chaining rule has a LHS logic that matches the goal requesting Rudolph s nose The RHS uses an OPS 2000 library function to ask Rudolph the question If Rudolph says yes then the goal is proven retracted otherwise the goal is refuted At last Santa has his headlight and Rudolph makes his break for fame Santa makes one last statement which is given in the below rule MerryChristmas defrule MerryChristmas christmas eve foggy weather headlight printf Merry Christmas Ho Ho Ho n www siliconvalleyone com Page 49 of 56
8. OPS 2000 User s Guide Advanced Use of Rule Sets example file wine ops The Wine Expert WinEx example program given in the OPS 2000 example set is a good example of the power of rule sets WinEx has nine rule sets Two of which are always active CombineCertainties and PhaseControlRules Ho PhaseControlRules CombineCertainties priority 100 priority 10000 state active state active 7 wine selection rulesets priority 0 state inactive Rule set CombineCertainties is given the highest priority in the system it is responsible for the combination of wine certainties This rule set utilizes a fuzzy rule to perform this task Rule set PhaseControlRules is given the lowest priority in the system its responsibility is to sequentially activate the seven rule sets responsible for the wine selection process Each rule set is allowed to run to completion and is then subsequently deactivated by the PhaseControlRules rule set which then activates the next rule set in the process When all of the wine selection rule sets have been run PhaseControlRules deactivates CombineCertainties and itself and then stops the WinEx expert object PhaseControlRules activates and deactivates the wine selection rule sets in the following order I activate AskQuest
9. Reasoning Environment cccccccceeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeseeeeeeeeeeeeeseneees 35 WORKING ENVIRONMENT ci a A 36 VIEWING INFORMATION Sii 38 MONITORING EXECUTION ccoo toc cccecaddn ciao a iodo dba doin cea 39 DRIBBLE saanen arado tey 50 40 Function GallS icon saai aniani aa aE a aana Saan aaiae aaaea Eak iea aaan 41 INTERPRETED PUNCTIONS 00 da 41 COMPILED FUNCTIONS cocina AA a aa eaii 42 Embedding OPS 200D isicicricininin ds 43 EMBEDDED APPLIGATION i sisccccsstecdscctenss die T EE AE RR A A 43 EMBEDDEDOPS 20005 000 li did tdi aii dd 44 Inference Engine Library oionionicinaianaii ai 45 Example Profa Moab 46 Advanced Use OF RUNG Seis usara 50 Advanced Use of Expert Objects 2 cccccccsesseeeeeeeeeeeeeeeeeeeeneeeeeeeeeeeeeeseeeeeeeeeeeeeeeeeseeseeeennees 51 Example Object Oriented Program miiocoioseonsancncncncncnncanancnanincne noo enana dae enana 52 OPS 2000 Help DESK innin inant teaa dis 55 www siliconvalleyone com Page 4 of 56 OPS 2000 User s Guide Introduction This user s manual is intended to serve as an introduction to the OPS 2000 system It covers the basic concepts behind the system and also gives some example programs Overview OPS 2000 is a hybrid interactive rule and object based software development environment The rule based interpreter provides facilities for both forward and backward chaining Where some languages half support either forward or backward chaining OPS 200
10. alumni He is accessible via e mail at frank siliconvalleyone com www siliconvalleyone com Page 5 of 56 OPS 2000 User s Guide Notes The design goal of OPS 2000 was to create a knowledge reasoning system from the point of view of a procedural object oriented programming language C The most obvious difference between OPS 2000 and its predecessors is that all rule variables are typed OPS 2000 also provides the concept of active and inactive rule sets with dynamic run time priorities and local agendas Lastly but most importantly OPS 2000 was designed and developed to run on sequential and parallel architectures Many traditional artificial intelligence people preach that the LISP notation for rule representation is best OPS5 and other rule based languages follow this philosophy However Silicon Valley One recognizes the large amounts of time knowledge engineers spend finding software errors that could have been caught by the compiler had variable type information and fact format information been specified Silicon Valley One also recognizes that OPS5 like languages tend to have knowledge reasoning and representation models that are too limited making them ill suited for developing complex systems Finally for those who enjoy the LISP style of rule representation we suggest taking a simple expert system problem and code it in a LISP like rule based language and in OPS 2000 For the most part you will find that those variable names that
11. an OPS 2000 inference engine function library that is accessible from both the interpreted and compiled environments www siliconvalleyone com Page 34 of 56 OPS 2000 User s Guide Knowledge Reasoning Environment The OPS 2000 operating environment has two primary features the C interpreter and the Current Working Environment CWE The CWE refers to where in the knowledge reasoning system the interpreter is currently focused The operating environment s tree like object structure has a root with its children being all of the currently loaded expert object definitions The root its expert objects and all of those expert objects rule sets each have their own working environments This structure can be traversed using the ce function ROOT expert object 1 oar expert object N PRSI i RSM This operating environment object hierarchy allows the developer to simultaneously debug and test multiple expert objects Many functions such as agenda and facts give information dependent upon what the current working environment is Variable declarations that are local to each working environment can be displayed using the list function It is recommended that you ce change environment to t
12. consequently resorted to cannibalism to get his thesis advisor off his back The dean whose name is Dean doesn t tolerate beer on campus but otherwise is human The students well are students they study twenty four hours a day and wander the campus pondering how to pay next semester s tuition bill The students also pick up anything they find such as pizzas magical wands and beers Each person starts off in their natural locations Grendell in the dungeon Dean in the dean s office and the students Charlene Frank and Karen in Charlene s office the Computer Laboratory and the Altgeld math building respectively Each person has a threshold level which is a measurement of the time interval required for the person to become so bored that they have to do something There is a system clock that all instances of people share This clock acts like father time for with each tick each person becomes a bit more restless When a person becomes so restless that their threshold level is reached they move When a person moves their restlessness is reset to zero In this program there are three types of people normal troll and dean Each type of person has a corresponding method of responding to a move message Consequently the move message is declared as virtual in class person A summary of each type of person and what happens when they move is given below Each summary is labelled with a person type an english description of the person s moves
13. data object matching a pattern condition u x is x s degree of membership Variable y is the membership value of a slot specification www siliconvalleyone com Page 27 of 56 OPS 2000 User s Guide Statistically Dependent A fuzzy rule defaults to sd Its rules are lt pattern condition gt y u x not x y 1 u x and x1 x2 x3 y min u x1 u x2 u x3 or x1 x2 x3 y max u x1 u x2 u x3 Statistically Independent This is sometimes referred to as possibilistic logic Its rules are lt pattern condition gt y u x not y y 1 u y and y1 y2 y3 y u y1 u y2 u y3 or y1 y2 y3 y u y1 u y2 u y3 u y1 u y2 u y3 www siliconvalleyone com Page 28 of 56 OPS 2000 User s Guide Backward Chaining A backward chaining rule has the generic syntax given below defrule lt name gt bc lt LHS gt lt lt RHS gt The LHS is a single goal pattern and the RHS is a subgoal list and or pattern logic The pattern logic is pattern matched exactly as though it were a forward chaining rule s LHS OPS 2000 has two types of working memories fact and goal The GWM differs from the FWM in that a goal s existence can be dependent upon the existence of other goals The purpose of a goal is to state a hypothesis that is to be proven or refuted Goals can either be of the free form or relation data for
14. fire remove makes sure that regardless of what happens to an activation when it fires that it is removed from the system after it has fired www siliconvalleyone com Page 45 of 56 OPS 2000 User s Guide Example Program example file xmas ops This example problem is given by creating rules that fit a story s script Our example is a christmas time problem The problem is that it is a foggy christmas eve and Santa Claus cannot lift off with his regular reindeer crew due to limited fog visibility The story begins with Santa pacing the floor frantically trying to think of a solution to his visibility problem Santa the genius he is decides that what he needs is a headlight So he makes a goal to find a suitable headlight for the heavy christmas eve fog Our example program will be a Santa expert object So we start by declaring an expert object with the name SantaExpert This is given below defeo SantaExpert Now let s think There are two primary components to a typical expert object definition These being a collection of rule sets and an initial state An initial state definition would have to encompass where the knowledge reasoning process starts The story began it is christmas eve and there is foggy weather defeo SantaExpert deffacts NorthPole Christmas eve foggy weather Santa Claus realizing he has a crisis on his hands logs onto the north pole s computer network Immediately recognizing t
15. gt OutputDevice gt cwe Display the CWE OutputDevice gt list rs OutputDevice s rule sets OutputDevice gt list vars OutputDevice s local variables OutputDevice gt agenda OutputDevice s primary agenda OutputDevice gt facts OutputDevice s FWM OutputDevice gt goals OutputDevice s GWM OutputDevice gt set cwe false gt Still in OutputDevice gt reset Reset OutputDevice gt run I Run OutputDevice gt cel Return to the root gt run 1 Run th ntire system gt www siliconvalleyone com Page 37 of 56 OPS 2000 User s Guide Viewing Information There is a set of system functions responsible for displaying system information A brief summary of some of the more useful system functions is given below Please see the reference manual for more exact information and also note that references to a current expert object or rule set are referring to when the current working environment is set to an expert object or rule set respectively agenda Displays the agenda of the current expert object or rule set facts Displays the fact working memory of the current expert object goals Displays the goal working memory of the current expert object fctns Displays all of the loaded C function prototypes channel Displays a channel s contents list Lists out various types of system information curre
16. integers and floating point numbers reals A fact has at least one element The basic format of a fact resembles that of a sentence a sequence of words However a fact cannot have any punctuation A free form fact has no predefined format constraints For example the following are all legal OPS 2000 facts 1 The monkey is at position 22 The car is red Frank is twenty four The shuttle has reached main engine cutoff Hurricane Gilbert missed Houston I am 99 9 percent sure of the problem Charles is twenty four The value is 314 0E 2 O IO OB w www siliconvalleyone com Page 14 of 56 OPS 2000 User s Guide Relation Fact A free form fact is treated by the compiler as a sequence of words with no constraints on its size or field types This may seem great at first however in use this usually leads to confusion and hard to find software errors Relation facts introduce format constraints fact properties and the flexibility of input output formats Four benefits of relation facts are given below 1 Predefined format constraints can be checked by the compiler 2 A relation fact is only pattern matched to patterns of the same relation 3 Predefined input output formats increase the flexibility and consistency of a knowledge base interface 4 Arelation fact can have predefined properties Fact relations are declared using the defrelation declaration which has the generic syntax given below defrelation lt name gt lt f
17. monkey can be asserted into the knowledge base A class object fact does not make a copy of the asserted class object but rather it contains a reference to its asserted class object Consequently a class object can have multiple references appearing in the knowledge base and in the local C interpreter environment Therefore some special attention must be given to the retraction of a class object from the knowledge base A retract statement can be used to remove a class object from the knowledge base However if the real intent is to delete the class object from OPS 2000 then a delete followed by a retract is required For more information see retract in the reference manual www siliconvalleyone com Page 18 of 56 OPS 2000 User s Guide Fact Preprocessor Any time a nonclass fact is asserted it is preprocessed to determine whether or not it is a relation or free form fact A relation fact is translated into its generic form relation gt generic A gu a Pee a form asserted 4H 2p fact gt preprocessor gt free form www siliconvalleyone com Page 19 of 56 Patterns OPS 2000 User s Guide Each of the three WME types have a corresponding pattern type A free form fact is only compared to free form fact patterns A relation fact is only compared to patterns of its relation s type La
18. production There are two types of condition elements usually found in the LHS of a production tests and patterns The test condition is usually a test for constraints on the pattern conditions that have been satisfied A pattern condition is satisfied when it matches a data object found in the working memory Once each element of a rule s LHS is satisfied the rule is said to be ready to fire An instantiation is an ordered pair of a production and the working memory elements that satisfied the production s LHS Each cycle of the production system s execution may produce one or more rules which are in the ready to fire state This set of satisfied rules is called the conflict set and the production system s conflict resolution algorithm determines which elements of the conflict set will be fired When a rule is fired its RHS is executed and it is removed from the conflict set This entire process is called the recognize act cycle The production system first recognizes the rules that are in the conflict set then uses the conflict resolution algorithm to choose instantiations from the conflict set and finally executes the RHS s of the productions associated with the selected instantiations In OPS 2000 there is a default recognize act cycle function and there are also facilities for user defined recognize act cycle functions Working Memory The Working Memory WM is a collection of data objects that represent the current state of the system A
19. production system s working memory is like a classroom blackboard Each production watches the blackboard for new information As new information data objects streams onto the board a production determines all of the possible ways its LHS can be satisfied For each possible way that the production s LHS can be satisfied an instantiation of the production is placed into the conflict set An instantiation is removed from the conflict set when it is fired or if any of the working memory elements that created the instance no longer exist The initial state of the working memory is defined by the user and is modified using a production s RHS operators Data objects are asserted into the working memory using an assert operator and are retracted from the working memory using a retract operator Interpreter The interpreter of a production system cycles through the recognize act cycle The interpreter must first match the contents of the working memory to each rule s LHS The interpreter creates the conflict set by finding all of the productions that have become instantiated from the current working memory Once all of the matching has been done and a conflict set created the interpreter uses the conflict resolution algorithm to determine the set of instantiations that it will fire Once these instantiations have been fired the interpreter repeats the cycle System execution terminates when after the match phase the conflict set is empty This is true be
20. 0 has both types of chaining built into its design The small C interpreter contains about 75 of the features of the C version 1 0 language specification A full interface to external compilers is provided so that compiled functions can be directly called from the C interpreter OPS 2000 compiles all code into an intermediate representation that it interprets Innovation The OPS 2000 system was the first knowledge engineering tool in the world to be designed for the C programming language it was the first commercially available C interpreter of any kind and it was the first C product for Microsoft Windows However OPS 2000 v2 1 s C interpreter is still a limited implementation of the C language An OPS 2000 rule s action is a special C compound statement that includes language extensions to support knowledge engineering assert statement refute statement printout statement Creation The OPS 2000 system was created and documented by Frank Lopez who is also the founder of Silicon Valley One Mr Lopez received his BSCS from Purdue University in 1986 and his MSCS from the University of Illinois in 1987 His major areas of study and research have included computer engineering operating system design computer language design parallel processing theory of computation object oriented programming and artificial intelligence He is a former NASA engineer AT amp T Bell Laboratories scientist Oracle alumni and Microsoft
21. MAE nd e a al te e a o O ee ce e e a A aaa 14 WORKING MEMORY ELEMENTS oococcccccccconononononncnnnnnnonnnononnnnncnanono no nn nnn cnn an anno nn nn n nn cana n nana rn nn cnn nana rn nara c anna rana 14 FIO O FOrin Fat bi Le eae 14 EA A a Pt A EA EA EARE 15 Binary Relativa aida 16 A A 17 CLASS QDJOCE af 6 AET AEAEE A A A A A aid 18 FACT PREPROCESSOR Saa AI ds 19 PATTERNS O A A AA AS ETETA 20 EFOB ROMA A Sh rae as oe aa eS te aT Te eo RR Bee vas een 20 FICIAUON RAC eee as aN eh te SASS IER NN SDE AE DE hes TETA Oe bd ade des ara Ll 21 Clas ODjJOCE FACE PEA veh A A A AA AAA 22 Gk ORR R EEE ERE EER EE CEE PEEP EEE REECE EET EEC EE EEE Ee CEE REE EET EEE TES CrEC EEE PELE EE EET REE REE c TE EEE EEE EES EEC TREE ERe CRE PEE EEE PEEP re reer ere 23 FORWARD GHAINING 32 22 cov da eee 24 CONTIDENCE Factor anre A ea aera BIND cP TO A IE 26 III TUF ices So cc o IS ON lh hd chk at Ma he oe oe oka eel aM onto dal colt eek le eS 26 MAMI roo cos 26 RUZ A A A pil 27 Statistically Depende and 28 Statistically Independent cido toi sorted pica iria 28 BACKWARD CHAINING outils id e io 29 Hule SOS iia AA A dad 31 LOCALAGENDA att Ne ride jected dnettietie 31 gt 32 KNOWLEDGE BASES a E ea 32 Fact Working Meno a ee 32 Goal Working MEMOIYV seiis siini At AAA A Suelo Ai 32 www siliconvalleyone com Page 3 of 56 OPS 2000 User s Guide INFERENCE GYGCLE servei lc lira lr parra 34 PRIMARY AGENDA A EAE A E lao dados cute nada 34 Knowledge
22. OPS 2000 User s Guide OPS 2000 User s Manual Version 2 1 Copyright c 1988 2012 by Silicon Valley One P O Box 77782 San Francisco California 94107 WWW SILICONVALLEYONE COM The United States of America All Rights Reserved www siliconvalleyone com Page 1 of 56 OPS 2000 User s Guide SILICON VALLEY ONE LICENSE AGREEMENT This is a legal agreement between you LICENSEE the end user and Silicon Valley One LICENSOR By opening this package you are agreeing to be bound by the terms of this agreement If you do not agree to the terms of this agreement then you must return this package and all related materials to Silicon Valley One for a refund within thirty 30 days from the date Silicon Valley One shipped this product to you The terms and conditions set forth on this page of the agreement shall apply to use by LICENSEE of the OPS 2000 Software Product herein referred to as PRODUCT DEFINITIONS APPLICATION means any instruction or instructions in source code or object code format referred to as binary format for controlling the operation of a CPU SOFTWARE PRODUCT means materials such as Applications information used or interpreted by Applications and documentation relating to the use of Applications OPS 2000 MATERIALS is defined as all PRODUCT documentation all PRODUCT distribution software all PRODUCT backups and this license agreement TERM LICENSEE may terminate its rights under this Agreemen
23. all i vall apple sells for 0 26 and we have 125 yes r vall 0 26 i vall 125 apple r vall i vall apple sells for 125 and we have 0 26 no www siliconvalleyone com Page 21 of 56 Class Object Fact Class patterns are used to match class object facts lt class name gt lt member attribute gt lt pattern pattern_element gt OPS 2000 User s Guide A class pattern can contain zero or more member attribute specifications The member attribute can be either matched to a pattern element or to a class pattern If it is a class pattern then the member attribute must be a pointer to the nested pattern s class The following C class will be used in the examples given below class bread symbol name real price integer quantity y symbol syml class pattern class object matches bread name syml bread name wonder yes syml wonder bread name country bread name country yes bread bread name generic yes A class pattern without any class attributes will match all class objects of that type www siliconvalleyone com Page 22 of 56 OPS 2000 User s Guide Rules There are two primary types of rules in OPS 2000 forward chaining and backward chaining Forward chaining rules can be further divided into two more groups those having threshold expressions and those that don t A rule type summary diagram is given below 4 rule
24. as a pizza take that instead of the person s life If the person has no pizza allow them to pray for sometimes praying will move them to a nearby location If the person has no pizza and praying didn t work and the person has a magical wand then the person can wave the wand to move to a random location However the wand only has a fixed number of charges and the random location may be the current location so the wand won t always save a person s life Lastly if none of these three things have saved the person s life eat the person and return to the dungeon for some rest 3 Otherwise randomly move from the current location www siliconvalleyone com Page 53 of 56 OPS 2000 User s Guide The Campus The pizza adventure game uses ten campus locations A diagram illustrating how these locations are connected is given below 4 4 Dean s CAB s Office Office UP UP DOWN DOWN EAST EAST English t gt Altgeld AR gt Computing Building t lt Building lt Applications WEST WEST UP NORTH NORTH DOWN SOUTH SOUTH EAST Computer Union gt Digital Comp Laboratory Building lt Laboratory 4 4
25. cause only the RHS of an instantiated production can modify the working memory www siliconvalleyone com Page 13 of 56 OPS 2000 User s Guide OPS 2000 primer Complexity There are four types of knowledge reasoning objects facts rules rule sets and expert objects 1 Facts represent the active data in a knowledge reasoning process Rules act upon facts and are grouped into sets Rule sets address rule level complexity Rule sets provide a means to package rules in a module that based on need can be added and removed from a knowledge reasoning process 4 Expert objects provide a means to define a knowledge reasoning process A knowledge reasoning system is composed of one or more knowledge reasoning processes expert objects that can share information via interobject channels Working Memory Elements In OPS 2000 working memory elements are often referred to as facts productions as rules and instantiations as activations The conflict set is stored in a data structure called an agenda There are three types of working memory elements free form fact relation fact and class object 4 working memory element 4 free form fact relation fact class object Free Form Fact A fact is composed of symbols words
26. clare has a string constant parameter www siliconvalleyone com Page 41 of 56 OPS 2000 User s Guide Compiled Functions Compiled C functions can be directly called from OPS 2000 There is an object file form of OPS 2000 included in the distribution disks This file is for the host environment s C compiler The way in which OPS 2000 calls your compiled functions is through an interface defined by function ops_def_fctn This function is given a pointer to your compiled function and an associated C function prototype OPS 2000 calls function user_fcins when it is started or reset This function is not defined in the OPS 2000 object file and must be defined by you All of your compiled function calls to OPS 2000 functions that set some state in the system should be performed from this function These include ie default ops def fctn ops declare ops free form The steps for adding a compiled function are outlined below 1 a fF ON Define your function and compile it into an object file for linking Place a ops_def_fctn declaration for your compiled function 1 in user_fctns Compile user_fctns into an object file for linking Link the OPS 2000 object file with the above two object files The new executable now includes your compiled function which can be directly called from your OPS 2000 program From the OPS 2000 system there is no difference in the calling conventions of compiled and interpreted fun
27. ctions www siliconvalleyone com Page 42 of 56 OPS 2000 User s Guide Embedding OPS 2000 OPS 2000 can be embedded within a larger piece of software and other systems can be embedded within it Consequently a particular version of OPS 2000 can be embedded within an application and simultaneously have applications embedded within it In terms of licensing regardless of how OPS 2000 is embedded in an application the application attached to OPS 2000 cannot be redistributed without an explicit embedded license agreement from Silicon Valley One Embedded Application Applications can be embedded within OPS 2000 using the library function ops_def_fctn This compiled function call binds an externally compiled function to an interpreted OPS 2000 C function prototype A feature of OPS 2000 s external function calls is that they can pass by reference single dimensional arrays of data of a fundamental type integer int real double char symbol char This allows arrays of data to be passed back and forth between OPS 2000 and external functions From within a rule a matching working memory element can be converted to its text representation and then subsequently passed as a string parameter A function parameter can also be used as a buffer to retrieve data from an external source For example Example defrule Example match var Svar lt matches any free form fact char buffer 120 Assert the
28. e age n RuleOne www siliconvalleyone com Page 25 of 56 OPS 2000 User s Guide Confidence Factor E Confidence Factor mn gt min cl c2 mx gt max cl c2 defrule lt name gt cf mn mx lt threshold expr gt lt pattern variable declarations gt lt pattern logic gt gt lt compound statement gt Description Minimum A confidence factor rule defaults to type mn Variable is set to the minimum confidence factor value of the match set Maximum Variable is set to the maximum confidence factor value of the match set The mx symbol is used to specify this type www siliconvalleyone com Page 26 of 56 Fuzzy OPS 2000 User s Guide Fuzzy rule sd gt statistically dependent si gt statistically independent ae lt name gt fz sd si lt threshold expr gt lt pattern variable declarations gt lt fuzzy pattern logic gt gt lt compound statement gt Description The fuzzy pattern logic can be any legal pattern logic However the pattern logic operators and or and not take on completely new meanings DeMorgan s theorem is not applied to a fuzzy rule s pattern logic The logic structure is not altered in any way Regardless of the logic each and every one of the rule s test and pattern conditions must be satisfied in order for the rule to be satisfied Variable x is defined to be a
29. e reloaded www siliconvalleyone com Page 9 of 56 OPS 2000 User s Guide Object Oriented Programming An object oriented programming system is one where programs are executed by sending messages between data objects Objects can only process messages that they have been defined to handle In a true object oriented system the internal structure of a data object is hidden from other objects in the system data encapsulation An object s internal values can only be modified by the object s operations Each data object has a type or class that specifies the object s data representation and the operations or messages that it can process A class can inherit properties from other classes inheritance C is an example of an object oriented programming language OPS 2000 implements an interpreted version of this language that includes all of its object oriented features except one form of overloading Function overloading has been implemented but operator overloading will appear in a future release A C Class Instance s Life Cycle 4 4 4 New Create Call Operator gt Class gt Appropriatet gt Instance Constructor 4 4 4 Program Uses Instance Destroy Call DELETE Class lt Destructor lt Operator lt
30. eclarations are asserted into their respective working memories using the reset function Function run is used to cycle the inference engine initial state reset do compare do knowledge aaa acer a rules base assert run place retract A EE S select activation F fire E SPSS S n lactivations activation agenda Primary Agenda An expert object s primary agenda is composed of each of its active rule sets nonempty local agendas Rule sets are ordered in this agenda based on the following rules in descending importance 1 Highest rule set priority 2 Highest rule activation priority 3 Most recent working memory element 4 Most recent time stamp Inference Engine Function An expert object s inference engine function is responsible for selecting a rule to be fired from the primary agenda This function can be explicitly specified for each expert object using the ie declaration There is a default inference engine function for expert object s that don t have one explicitly specified This function will fire and remove the activation that appears at the front of the primary agenda The default inference engine function can be directly specified using ie_default Function ie_default can only be called from an externally compiled function There is
31. ed using function list gt gt list eo gt In the above example function list will display the names and states of the currently loaded expert objects To run this system we first must reset the entire system When function reset is executed at the root environment it will reset all of the loaded expert objects It is necessary to reset a loaded expert object because when it is loaded it is initially placed in the stopped state Function reset places an expert object in the running state gt gt reset gt This example must be run from expert object OutputDevice due to its data driven design Running expert object OutputDevice can be accomplished in two ways gt run OutputDevice 1 from the root environment gt reset gt ce OutputDevice OutputDevice gt run 1 from OutputDevice s environment OutputDevice gt www siliconvalleyone com Page 36 of 56 OPS 2000 User s Guide The prompt string OutputDevice is the default prompt prefix when within expert object OutputDevice The symbol appearing between the opening and closing brackets indicates the current working environment Function cwe also will display this same string value This CWE prefix prompt can be turned off by using the function call set cwe false Function env can be used to display the value of the current working environment OutputDevice
32. f 56 OPS 2000 User s Guide Tutorial Getting Started This tutorial will cover some of the basic features of the system Your first task is to startup the OPS 2000 system This can be done by typing ops This places you in the OPS 2000 Operating Environment OE This environment is aimed to be the C equivalent of a Lisp interpreter This environment has shell variables editor prompt local variables C globals and program definitions The prompt gt indicates that the OE is ready to accept keyboard input In this tutorial this prompt serves as an indication to you that you should enter into the OE any information appearing immediately after it The OE begins processing a statement only after a carriage return has been entered The OE treats everything entered at the prompt as a C statement Each entered statement is compiled and then executed The OE interpreter views compiled definitions from the outer most program level which is where global variables are declared Consequently global variables can be directly declared and manipulated from the OE prompt In addition the OE has a set of shell variables that are used by its internal functions for such values as the OE prompt string and the current editor name used by function edit Shell variables can be displayed using the env function and set using the set function A summary of the shell variables is given in env s reference manual description Now for our
33. fied by the list of subgoal specifications that appear between the parenthesis in this rule s RHS Here the group only has one subgoal specification If this subgoal group is proven then so is the headlight goal Likewise since only one subgoal group will be spawned from this goal then if this subgoal group is refuted then so is the headlight goal The Candle Maker Train Tooter and Glitter Gluer all receive the Chief Helper s message The Candle Maker proposes a big candle CandleMaker defrule CandleMaker bc bright light lt big candle The Train Tooter proposes a toy train TrainTooter defrule TrainTooter bc bright light lt toy train Lastly the Glitter Gluer proposes Rudolph s nose www siliconvalleyone com Page 47 of 56 GlitterGluer defrule GlitterGluer bright light lt rudolphs nose OPS 2000 User s Guide The Chief Helper takes a look at his subordinates recommendations and decides that the best solution is Rudolph s nose First we give a rule that says if Rudolph has previously agreed that Santa can use his nose then the goal is proven This is to avoid asking poor Rudolph questions that he has previously answered RudolphsNose defrule RudolphsNose priority 100 rudolphs nose rudolphs nose This is matched to the GWM This is matched to the FWM The above rule has a
34. first example gt printf Hello World n The system prints out Hello World Function printf is the actual C printf function from the standard C input output library A semicolon is automatically inserted at the end of each entered statement Our next example will show some of the OE s features gt fctns Function fetns lists all of the C function prototypes for all of the currently loaded C functions Any of these functions can be directly called from the OE These functions can either be user or system defined Function fetns is an example of a compiled C system function The user can add as many functions as needed but the user cannot specify a function prototype that is already bound to another function definition This binding definition can either be bound to a compiled or to an interpreted function This binding is completely transparent to the user Function declare is provided so that declarations can be made at the OE prompt Let s add three integer variables to the OE Variables entered at the OE prompt are treated as C globals gt declare int a b c Function declare expects to have complete C declarations Now let s give these variables some values gt a 10 b 20 c a b Let s use printf to check these variables values gt printf a Sd b Sd c Sd n a b Cc Now let s try some loop control gt for a 0 a lt 10 a printf OPS 2000 R
35. haining rules with pattern logic then its parent goal is refuted If a backward chaining rule has pattern logic then any goal that matches the rule s goal pattern will have an anonymous subgoal group that exists for the life of the goal This is due to the fact that the pattern logic becomes an active participant in proving the matching goal For example www siliconvalleyone com Page 29 of 56 OPS 2000 User s Guide JK Goal Proven by FWM Une defrule Goal Proven by FWM bc segment value value goal pattern matches GWM value pattern logic matches FWM Goal_Proven by FWM This rule states that any goal that has a matching fact is proven Consequently any goal that matches the above goal pattern will exist until it is either proven by a subgoal group proven by pattern logic or is explicitly removed by the application www siliconvalleyone com Page 30 of 56 OPS 2000 User s Guide Rule Sets The purpose of a rule set is to reduce knowledge base complexity A rule set is a collection of rules that are grouped together as a module A rule set module can then be activated and deactivated An activated rule set is treated as an active element of the knowledge reasoning process A deactivated rule set is treated as though it didn t exist When an active rule set is deactivated its local agenda is removed from its expert object s primary agenda Thus at runtime a rule set s state can fluctuate between be
36. he expert object you wish to run A typical sequence for loading and running a particular expert object is given below gt gt load mab ops gt ce MFE gt reset gt un 1 gt MFE is the name of the expert object defined in file mab ops www siliconvalleyone com Page 35 of 56 OPS 2000 User s Guide Working Environment The OPS 2000 interpreter has a current working environment variable This variable is set to the object in the system with focus A object is given focus using the ce function Function cwe displays the value of the CWE variable Why a current working environment Well this allows focus to be set on a particular object so that it can be directly examined This object can be either an expert object or a particular expert object s rule set For example let s load up an example problem and show how this works When you first start OPS 2000 the system places you at the root working environment Anytime you want to return to this environment from another environment simply use the ce function with no parameters Here we use function clear to remove any previously loaded definitions gt gt ce Returns you to the root gt clear Clears anything that has been loaded gt Now we will load up example problem sys ops gt gt load sys ops gt Now the system has eight expert objects The names of the loaded expert objects can be display
37. he power of electronic mail e mail he creates a message saying that he needs a headlight So with the touch of a key he sends his message across ICE NET to all of his fellow north pole hackers The following rule will do the trick defrule SantaClaus christmas eve T1f it is christmas eve foggy weather and foggy weather not headlight and Santa doesn t have a headlight then assert a goal to get Santa a headlight assert headlight gt GWM Here the LHS of the rule describes when the rule should become activated When the rule fires the goal headlight will be asserted into SantaExpert s GWM www siliconvalleyone com Page 46 of 56 OPS 2000 User s Guide Luckily the Chief Helper was logged on exactly when Santa sent his message Upon reading the message he decides that what is needed is a bright light So he sends a message to his subordinates requesting ideas for a bright light Given below is a backward chaining rule that reflects the Chief Helper s logic ChiefHelper defrule ChiefHelper bc headlight lt bright light The rule given above is a backward chaining rule so it has the type specifier bc A backward chaining rule has a LHS goal pattern and a RHS that can be a subgoal list and or pattern logic When this rule fires a subgoal group will be spawned from the headlight goal which becomes the parent goal of the subgoal group This subgoal group is speci
38. ield specifiers gt lt field types gt lt priority specifier gt lt property specifiers gt interface input output lt name gt lt string constant gt A relation s generic form has the following format lt relation name gt lt field specifiers gt The generic form is a relation fact s default input output format www siliconvalleyone com Page 15 of 56 OPS 2000 User s Guide Binary Relation Binary relations can have zero or more of the following properties reflexive transitive and symmetric For example defrelation brother of brotherl brother2 symbol brotherl brother2 property transitive 1 brother of eddie jimmy 2 brother of jimmy michael implies 3 brother of eddie michael If the first two facts exist but the third doesn t then the system will create a rule activation that will assert the third fact Relation facts take the form of the relation s generic form which is defined by the sequence of variable types given in the parentheses immediately after the relation s name When the three example facts are compiled they will produce the following variable bindings 1 brother of eddie Jimmy brotherl eddie brother2 jimmy 2 brother of jimmy michael brotherl jimmy brother2 michael Priority A relation definition can include a priority declaration To understand why this declaration exists one must first understand how a relation s
39. ing an active or inactive element of a knowledge reasoning process as the need for its knowledge varies A rule set is composed of one or more rules a local state a priority and a local agenda A rule set declaration cannot be nested and can only appear within an expert object s definition making the definition local to a particular expert object 4 4 rules state 4 4 priority local agenda 4 4 rule set components A rule set s state is used to control whether or not its member rules are pattern matched to its expert object s knowledge base If a rule set s state is active then its rules are pattern matched Facts are not pattern matched to an inactive rule set A rule set s state can be given a default value by specifying a state declaration This value is reset each time an expert object is reset A rule set s runtime state can be changed using the activate and deactivate statements A rule set s priority is used to position its nonempty local agenda within its enclosing expert object s agenda This priority value can be changed at runtime using the activate statement All of a rule set s activations are placed into its own local agenda A rule set s nonempty age
40. ions 2 deactivate AskQuestions activate ChooseWineQualities Biz deactivate ChooseWineQualities activat RecommendWineQualities 4 deactivat RecommendWineQualities activate DefaultWineQualities Dis deactivate DefaultWineQualities activat SelectWines 6 deactivat SelectWines activate RemovePoorWineChoices Ta deactivate RemovePoorWineChoices activate PrintWines 8 deactivate PrintWines deactivate PhaseControlRules deactivate CombineCertainties stop WinEx www siliconvalleyone com Page 50 of 56 OPS 2000 User s Guide Advanced Use of Expert Objects example file sys ops This example uses a set of expert objects to simulate a computer circuit This circuit is composed of two multipliers an adder and an output device The adder and the multipliers each take two inputs The output device only takes one input Each of the boxes in the below diagram corresponds to an expert object The lines between the boxes represent interobject communication channels output device adder gt lt multiplier multiplier gt lt gt lt input input input input one two three four The input units feed input sets to the multipliers The multipliers feed their resul
41. is function calls the editor name stored in the OE s editor variable The OE has a set of special variables whose values can be set using the set function The function env displays these variables with their associated values Function load is used to load a file of OPS 2000 definitions into the system and function list can be used to list out the various types of objects that have been loaded For example gt load xmas ops gt list eo This example assumes that program xmas ops is in the current working directory In this example function load will first load xmas ops into OPS 2000 Next function list will display what expert objects are currently loaded There are many other string values that can be given to function list For example gt list fctns List out global function prototypes gt list vars List out the global variables Function clear clears and resets the system For example if an error occurs during a load it may be necessary to clear the environment of any objects that have been loaded Function load feeds a file s contents to the OPS 2000 interpreter when an error occurs loading terminates However any definitions that had been successfully loaded before the error occurred will still remain in the system Consequently since an existing definition cannot be overwritten by a new definition function clear may be necessary to reset the system so that the file can b
42. jects can be loaded into the system at once However to avoid problems it is recommended that you load all relation definitions before any expert objects are loaded If an error occurs when loading a file and the file is to be reloaded again function clear has to be used if any file definitions were successfully compiled before the error occurred This is due to the fact that any definition that is successfully loaded into the environment exists until the environment is cleared gensym A symbol created by gensym exists in the system forever Function setgen provides a manner in which you can recycle gensym values For example setgen 10 gensym Creates symbol gen10 gensym Creates symbol gen11 gensym Creates symbol gen12 setgen 10 gensym Reuses symbol gen10 gensym Reuses symbol gen11 gensym Reuses symbol gen12 In the above example gensym is called six times but only three symbol objects are created If you are constantly resetting your system you may want to use setgen in a initializing rule or function You should be aware of the fact that if you use setgen and more than one expert system is loaded into the OPS 2000 environment that it will affect the entire system www siliconvalleyone com Page 55 of 56 OPS 2000 User s Guide Class Object Members Resetting an expert object s knowledge base doesn t delete its class object members These members mus
43. king memory blackboard data memory current state knowledge base and an interpreter inference engine There are two primary types of production systems data driven forward chaining and goal directed backward chaining Both types of systems have many variations including some systems that do both types of inferencing such as OPS 2000 Data Driven A data driven production system uses the contents of the knowledge base to determine which productions can be executed Therefore by having the knowledge base describe the current state of some process the productions can then infer new states from the existing state In an expert system these states describe the current state of the reasoning process and the state transitions represent the application of an expert s expertise to the knowledge base In reality we are taking a given situation and applying the rules we know pertain to it Each time we apply a set of rules we come up with a new state Eventually we either draw some valid conclusions or we determine that the current state is based on incorrect assumptions and therefore is invalid The creation of multiple current states allows one to choose the best state at any given instance of time and thus stop the inferencing on any states that are known to be wrong In forward chaining we say this is what know now what conclusions can draw from this information So if there are quite a few possible conclusions but only one is really necessar
44. mat A goal can have one or more subgoal groups Subgoal groups are only generated by the firing of backward chaining rules The firing of a backward chaining rule can create at most one subgoal group All asserted goals regardless of their value are entered directly into the GWM When a goal is asserted into a GWM it is pattern matched to the appropriate patterns If a goal object causes a backward chaining rule to fire then this will either prove the goal or spawn a subgoal group Multiple rule activations can be generated by a single goal object which when fired can generate multiple subgoal groups A goal can be proven in two ways subgoal group or pattern logic If a subgoal group has each and every member proven retracted but not refuted then its parent goal is proven Pattern logic can prove a particular goal when that goal matches a backward chaining rule that has pattern logic in its RHS In this case when the pattern logic is satisfied such that its bindings match those of the goal s LHS pattern match then the matching goal is proven If a goal is proven then it is retracted from the GWM If a proven goal has no parent goal then it is asserted into the FWM If a goal is refuted then 1 It and all of its subgoals are retracted from the GWM 2 If it is a member of a subgoal group then that subgoal group is refuted If that subgoal group was the last subgoal group of its parent goal and that parent goal has not matched any backward c
45. matching fact into a string buffer assert var gt buffer Pass the buffer to an externally compiled function printf ss n buffer A function can retrieve a buffer of data MyData 120 buffer The character buffer can then be asserted assert buffer gt GWM assert buffer gt FWM assert buffer Mickey buffer EWM Example www siliconvalleyone com Page 43 of 56 OPS 2000 User s Guide Embedded OPS 2000 The distribution disks contains a special form of OPS 2000 that can be embedded within another application This special form contains a set of functions that allows an application to manipulate the OPS 2000 system A description of these functions appears in the reference manual as the embeddable library An example application is given below include lt stdio h gt include ops2000 h needed to run under windows OPS 2000 Application code to be compiled and linked with the x OPS 2000 embeddable object files E int OPS 2000 int i 2 while i gt 0 printf n n printf tWelcome to Indiana n printf in n ops init initialize the OPS 2000 system ops load dilemma ops load up Farmer Brown s Dilemma ops ce FarmerBrown change environment to FarmerBrown ops reset reset th xpert object ops_ run 1 run till stopped ops terminate terminate the OPS 2000 sys
46. nda is placed in its parent expert object s primary agenda Local Agenda A rule set s local agenda is composed of its rules activations A rule activation is placed into its rule set s local agenda based on the following rules in descending importance 1 Highest rule activation priority 2 Most recent working memory element 3 Most recent time stamp If a rule set is active then its nonempty agenda will be positioned in its expert object s primary agenda www siliconvalleyone com Page 31 of 56 OPS 2000 User s Guide Expert Objects An expert object is a knowledge reasoning process composed of one or more rule sets a state a primary agenda an inference engine function a knowledge base and zero or more interobject communication channels The purpose of an expert object is to create a process level partition in a knowledge reasoning system s complexity Most of today s expert system s can be encoded efficiently using a single expert object expert object components Function reset is used to reset an expert object which sets its state to be running The stop statement can be used by a rule to stop its enclosing expert object The state flag is only meaningful if the expert object s inference engine function interprets its value The system s default inference engine function will terminate execution when this flag is set The OPS 2000 operating environment provides facilities to simultaneously run multi
47. nt expert object statistics channel names expert object names scope variables rule names and rule set names www siliconvalleyone com Page 38 of 56 OPS 2000 User s Guide Monitoring Execution OPS 2000 has a watch trace mode which enables a developer to verify and debug code This watching mode enables the monitoring of function calls and returns working memory element assertions and retractions rule activations and deactivations rule activation firings inter expert object message sending and receiving the compiling and clearing of rules and functions and the activities of the inference engine The OPS 2000 system monitoring functions are watch and unwatch For example gt watch facts monitor WME assertions and retractions gt watch fctns monitor calls to and returns from functions gt watch mesgs watch inter expert object message passing A complete description of these functions appears in the OPS 2000 reference manual www siliconvalleyone com Page 39 of 56 OPS 2000 User s Guide Dribble dribble to flow in an unsteady stream The system function goals displays the current value of an expert object s goal working memory Function goals is often used when debugging a backward chaining application However sometimes it becomes necessary to record the system s output This can be accomplished using open_dribble and close_dribble Function open_dribble opens a dribble file
48. ple expert objects running set Information can be sent between the members of a running set via interobject communication channels Knowledge Base An expert object s knowledge base has two working memories fact and goal working memories Fact Working Memory The Fact Working Memory FWM is used to store facts for the forward chaining process Each FWM element s existence is not directly dependent upon any other FWM element s existence The deffacts declaration is used to specify this working memory s initial state The fact initial_fact is always asserted into the FWM when an expert object is reset Goal Working Memory The Goal Working Memory GWM is used to store goals for the backward chaining process A member of this working memory can have zero or more subgoal groups Each subgoal group is composed of one or more goals The defgoals declaration is used to create this working memory s initial state The goal initial_goal is always www siliconvalleyone com Page 32 of 56 OPS 2000 User s Guide asserted into the GWM when an expert object is reset www siliconvalleyone com Page 33 of 56 OPS 2000 User s Guide Inference Cycle The inference cycle has three basic components 1 Compare each fact to the applicable rules 2 Place any rule activations into the agenda 3 Select an activation to be fired goto 1 The knowledge base is initialized using deffacts and defgoals declarations The facts and goals within these d
49. properties are enforced When a binary relation has a property declaration system defined rule sets are created to enforce the properties that appear in the declaration A property enforcing rule set acts identically to a user defined rule set Consequently a relation s properties are enforced based on when the activations of a property enforcing rule set s rules are fired Thus a relation s priority declaration is actually a priority declaration for any system defined rule sets that are created for the relation The advantage of having a relation priority is that if a relation s properties are critical to the knowledge reasoning process then the relation can be given a very high priority and if the relation s properties are defined to help the knowledge reasoning process when no other knowledge can be applied then the relation can be given a very low priority www siliconvalleyone com Page 16 of 56 OPS 2000 User s Guide Interface Relation facts can also be given input output formats For example defrelation brother of brotherl brother2 symbol brotherl brother2 priority 100 property transitive interface input inl brotherl is the brother of brother2 input in2 brotherl and brother2 are brothers output outl brotherl is the brother of brother2 Here if a fact is asserted into the knowledge base that has either the format of in1 or in2 then the fact is translated to brother_of s generic format Fo
50. r example 1 Steve is the brother of David 2 Steve and David are brothers 3 brother of Steve David Given the above relation each of these three facts are identical This is due to the fact that both fact 1 and fact 2 are translated to the generic form while fact 3 is already in the relation s generic form Thus if all three are asserted at once and fact comparison is turned on see le eo _ wm_cmp only the first assertion will appear in the knowledge base A relation s input formats are global and can only be defined once Input formats are compared based on their variables types and not their variables names When defining an input relation each of the relation s variables must be used exactly once The system declares a generic input output format for each relation A limited form of a natural language interface can be created by declaring relations to have many possible input formats and by defining all system responses in terms of defrelation output formats This would allow for input and output formats to be modified system wide by simply modifying a set of relation declarations www siliconvalleyone com Page 17 of 56 OPS 2000 User s Guide Class Object Fact A fact can also be a C class object For example class monkey symbol name Monkey s name symbol on What monkey is on class object holds Object the monkey holds class location at Location of monkey he An instance of class
51. r interpretation of what we visually see the actual interpretation of the rule must be done by the knowledge reasoning system However if knowledge base complexity is kept in check this approach to problem solving can be very useful www siliconvalleyone com Page 24 of 56 OPS 2000 User s Guide A set of logical pattern operators can be used in a forward chaining rule s pattern logic these are and or and not The generic form for using these operators is lt operator gt lt slot specification gt A slot specification can be a pattern condition test condition or pattern operator The not operator checks for nonexistence and takes exactly one slot specification There is always an implicit logical and around all declared pattern logic For example Houston e defrule Houston houston is a city houston is in texas houston is oil gt printf Houston is one big highway n Houston internally is represented as Houston Ke defrule Houston and houston is a city houston is in texas houston is oil gt printf Houston is one big highway n Houston Test conditions are for testing interpattern values RuleOne X defrule RuleOn integer x y Jerry is x years old Carlos is y years old test x y gt rule variables pattern condition pattern condition test condition printf Jerry and Carlos are not the sam
52. rmal way humans go about problem solving SaveBaby defrule SaveBaby the baby is crawling across the road a convoy of Mack Trucks is approaching assert Get the baby out of danger This production could be a member of a set of productions that all dealt with protecting a baby from potential dangers By giving each rule a priority the system can determine which rule has the most certainty of giving the correct decision A major problem of current expert systems technology is dealing with the complexity of large systems of rules that simultaneously interact on the same problem OPS 2000 helps deal with complexity by adding several levels of modularity to the problem solving paradigm expert objects with interobject channels rule sets with local agendas and name spaces classes and relation facts In OPS 2000 an expert system can be composed of many subexperts that are each an expert at a subproblem Each of these experts can then be subdivided and with each new subdivision the system s complexity becomes more manageable In addition relation facts and C classes both improve the management of knowledge base complexity and verification www siliconvalleyone com Page 11 of 56 OPS 2000 User s Guide Production Systems An expert system encoded as a production rule based system is usually made up of three parts a set of productions rules held in production memory a set of data objects assertions held in wor
53. s best features is its ability to transparently call compiled and interpreted functions A function call such as fctns can either be bound to a compiled or interpreted function definition A function s binding is completely transparent to the caller function call binding interpreted compiled A function call is bound to a particular function definition binding This binding is completely transparent to the actual function call specification There are constraints on the types of objects that can be passed to and returned from compiled function definitions Interpreted Functions An interpreted function provides a simple and direct way to add function definitions to an OPS 2000 program An interpreted function can either be global system wide use or local to a particular class definition as are class operations constructors and destructors All class function members must have interpreted bindings Interpreted function definitions are loaded into the system using declare and load Function declare has limited use when called from the interpreter which is partially due to fact that there is a system limit on the size of string constants For example gt declare integer x print 10 The above parameter to de
54. st notify Silicon Valley One within thirty days of the date of transfer and you must indicate the name and address of the party to whom this software is transferred You may not disassemble or reverse engineer this software You may not modify this software You may not embed OPS 2000 in an Application if in any way it violates the nature of this agreement LIMITED WARRANTY This program instruction manual and reference materials are sold as is without warranty as to their performance merchantability or fitness for any particular purpose The entire risk as to the results and performance of this program is assumed by you However to the original purchaser only Silicon Valley One warrants the magnetic diskette on which the program is recorded to be free from defects in materials and faulty workmanship under normal use for a period of thirty days from the date of purchase If during this thirty day period the diskette should become defective it may be returned to Silicon Valley One for a replacement without charge Your sole and exclusive remedy in the event of a defect is expressly limited to replacement of the diskette as provided above If failure of a diskette has resulted from accident or abuse Silicon Valley One shall have no responsibility to replace the diskette under the terms of this limited warranty Any implied warranties relating to the diskette including any implied warranties of merchantability and fitness for a particular
55. stly a class object is only compared to patterns of the object s class Free Form Fact Fact patterns are used to match free form facts The generic syntax is lt pattern lement gt The below table gives some examples of free form fact pattern matching real r vall fact pattern fact object matches Houston is in Texas houston is in texas life is a dream integer 1 real r vall 140 250 35 0 10 240 3 0 Houston is in Texas Houston is in Texas life is yesterday integer 1 real 90 0 1 2 3 e0 20 3 www siliconvalleyone com yes no no yes r vall 90 0 no Page 20 of 56 OPS 2000 User s Guide Relation Fact Relation patterns are used to match relation facts The generic syntax is lt relation name gt lt pattern element gt A relation pattern s form must match in type and quantity that of the relation s generic form The below table gives some examples of relation fact pattern matching defrelation product name price quantity symbol name real price integer quantity interface input inl name sells for price and we have quantity input in2 name quantity price real r vall integer i vall relation pattern relation fact matches apple r vall i vall apple 0 26 125 yes r vall 0 26 21 vall 125 apple r vall i vall apple 125 0 26 yes r vall 0 26 21 vall 125 apple r v
56. t be explicitly deleted Two possible methods are given below Method 1 Method 1 requires that a special rule set be declared such that when it is activated it will retract all class objects While within an expert object environment the following example would work gt set rule true gt activate CleanUp gt run mk hg gt reset When variable rule is set to true the command line interpreter parses statements as if they were appearing within a rule s RHS The activate statement will activate the rule set s rules which will then subsequently be pattern matched to any class objects The rule given below would be useful in a cleanup rule set so that when all class objects have been retracted it will stop the expert object de gt frule EndCleanUp priority 10000 Lowest priority in rule set not 3 14 Some impossible fact stop EndCleanUp An example cleanup rule for class X is given below qd Method 2 frule CleanObject match object object lt X delete object retract object Method 2 requires that you define each class in a manner such that its constructors and destructors keep track of all asserted class objects www siliconvalleyone com Page 56 of 56
57. t by written notice to Silicon Valley One certifying that LICENSEE has discontinued use of and returned or destroyed all copies of PRODUCT If LICENSEE fails to fulfill one or more of its obligations under this Agreement Silicon Valley One may upon its election and in addition to any other remedies that it may have at any time terminate all of the rights granted by it hereunder by not less that two 2 months written notice to LICENSEE specifying any such breach unless within the period of such notice all breaches specified herein shall have been remedied upon such termination LICENSEE shall immediately discontinue use and return all copies of PRODUCT subject to this Agreement SOFTWARE LICENSE This software is protected by both United States copyright law and international treaty provisions This program and all related materials are to be treated exactly as though they were a book There is one exception one copy of the diskettes can be made for the sole purpose of backing up your software The running of this software analogous to the reading of a book can only be occurring on at most one computer at any given instance in time You cannot transfer this license unless it is done so permanently and with all OPS 2000 Materials You may transfer this license permanently by embedding OPS 2000 in an Application and permanently transferring that Application to a single party accompanied by all OPS 2000 Materials When transferring this license you mu
58. tem printf n n while return 0 OPS 2000 www siliconvalleyone com Page 44 of 56 OPS 2000 User s Guide Inference Engine Library The purpose of the inference engine library is to provide a set of functions that can be used to construct an Expert Object s inference engine function There is an interpreter binding for each member of the inference engine library This means an expert object s inference engine function can be either interpreted or compiled the former providing flexibility and the latter providing speed Among other things an explanation facility can be created using these functions For example function ie_act_rule_info gives a rule s summary line as one of its return values Thus each rule can have a declared summary line that explains its actions The inference engine function can then either store or display these lines as each rule activation fires Function ie_eo_stats returns an expert object s runtime statistics An activation can be fired multiple times before it is removed from the conflict set This can be done using function ie cs fire in conjunction with either ie_cs remove orie_cs fire _remove Please note that using ie_cs_fire on an activation followed immediately by ie_cs_remove on the same activation is not the same as using ie_cs fire remove on that activation This is due to the fact that when an activation fires its position in the agenda may change Thus ie_cs
59. ts to the adder s inputs and the adder sends its results to the output device A feature of the receive statement s current implementation allows for this system to be run by changing environment to that of the output device s expert object and then running it The receive statement will cause demand driven input to each of the expert objects from the four input units to the output device When the four input units stop producing data the multipliers will stop Once the multipliers have stopped the adder will stop which stops the output device www siliconvalleyone com Page 51 of 56 OPS 2000 User s Guide Example Object Oriented Program example file pizza ops Adventure Game This example program uses no public class data members This program was modelled after a Scheme program given as a class assignment at the University of Illinois Scheme is a Lisp dialect This adventure game takes place on a large midwestern campus The game uses ten locations three students a dean and a misplaced troll As with every campus there are things randomly distributed about it such as beers pizzas and magical wands the significance of each thing will be explained below You may be asking yourself the question what makes this an adventure game Well this misplaced troll whose name is Grendell is really a pizza loving cannibalistic person There are rumors that Grendell is really a graduate student who spent one to many years writing his thesis and
60. ules n www siliconvalleyone com Page 8 of 56 OPS 2000 User s Guide This loop prints OPS 2000 Rules ten times Now let s try some advanced expressions gt a b b gt 10 10 5 This is a C conditional expression The new value of variable a should be 200 There is a set of print functions that can be used to display an object s value gt print a The next example will first be given and then explained gt declare real abs real value return value gt 0 0 value value Pea gt gt printf Absolute values f SfAn abs 44 0 abs 55 0 Absolute values 44 0 55 0 gt In the above example we used function declare to enter a function definition The back slash character is used to continue an input line on the next screen line and it is immediately followed by a carriage return Function abs can now be used like any other OE function This manner of entering a definition limits its length to that of the maximum string constant length Another useful function is system This function takes a string parameter and treats it as a command line input to the resident operating system gt system vi text ops This example executes the vi editor with the file name text ops When the command has completed execution it returns to the OE For the purpose of editing a special function has been created that does the equivalent of the above operation edit Th
61. xperts Human experts tend to explain how they solve problems using heuristics rules of thumb Therefore if a computer system could learn to use these same rules then it would be considered just as much an expert as any of the human experts its knowledge came from These rules abstractly have the form if lt something is true gt then lt perform some actions gt and are called productions An expert system that is fabricated using a production system codes the heuristics of an expert domain into productions These rules are applied to the current state of the decision making process The input to the system consists mostly of data objects that describe changes to the current state of the reasoning process Once the system knows which rules have their conditions satisfied it has to decide which of these to apply to the current state to generate the next state of the reasoning process The actions associated with the selected rule s are executed each time a rule is applied By coding an expert s knowledge into productions we are able to easily update and modify any rules of the system Furthermore if the computer system gives a bad response or a response that could have been better it becomes a relatively easy task to locate the set of rules that led to the incorrect decision As can be seen in the example OPS 2000 production given below the heuristics the system uses have a declarative and modular representation which lends itself to the no
62. y this exhaustive searching may be unnecessary Goal Directed A goal directed production system backward chains the inferencing process This is done by assuming that a goal is true and then attempting to reverse engineer the application of the heuristics of the expert domain If all the knowledge exists in the system to support that a given asserted goal is true then the goal becomes one of the possible solutions of the system This can be seen as a guessing algorithm since we guess what the answer is goal and then attempt to prove that it is a valid solution In reality this is equivalent to saying this is what think happened now can it be supported By backward chaining the new states with known data and rules we are able to rule out states that cannot be supported by the knowledge base So if there are quite a few possible conclusions but only one is needed then this provides an avenue to contain the search from a fixed set of conclusions www siliconvalleyone com Page 12 of 56 OPS 2000 User s Guide Components of a Production System A production system has three main components productions working memory and an interpreter Productions The first component of a production system is the set of productions A production is a condition action construct Each production has a name LHS Left Hand Side antecedent condition and a RHS Right Hand Side consequent action The LHS is the logical AND of the conditional elements of the
63. you mistyped those fact formats that were missing a field and those data fields that you had in the wrong order will all be caught by the OPS 2000 compiler In addition you should find OPS 2000 s interface to external compilers to be one of the best in the industry Silicon Valley One is committed to bringing the state of the art in expert systems and object oriented technologies to the people Silicon Valley One welcomes your comments Correspondence should be sent to the address given below Silicon Valley One Attn OPS 2000 Users Group P O Box 77782 San Francisco CA 94107 www siliconvalleyone com Page 6 of 56 OPS 2000 User s Guide Manual Notation This manual uses various types of notation Symbols are used to describe a command s syntax The table given below describes these symbols Symbol Use symbol The bold font indicates a literal value that should be entered actually as it appears lt item gt Angle brackets indicates that you must enter the enclosed item s value symbol symbol The pipe serves as an associative or operation This indicates that exactly one of the or s values is to be entered symbol Square brackets indicates that the symbol s value is optional symbol The asterisk indicates that zero or more of the symbol s values is to be entered symbol The plus indicates that one or more of the symbol s values is to be entered www siliconvalleyone com Page 7 o

OPS-2000 Version 2.1 - User`s Manual

Contents

Download Pdf Manuals

Related Search

Related Contents