AI Unit 2

You might also like

Download as pdf
Download as pdf
You are on page 1of 22
LEARNING Leaming is the improvement of performance with experience over time, Learning Jementis the portion ofa learning AI system that decides hove to modify the performance pa and implements those modifications, d According to Herbert Simon, learning denotes changes in a system that enable a system to do the same task more efficiently the next time, Arthur Samuel stated that, “Machine learning isthe subfi ield of computer science, that gives computers the ability to leam without being explicitly programmed ”, In 1997, Mitchell proposed that, ” puter program is said to learn from “T’ and performance measure ‘P’, if its as measured by ‘P’, improves with experience E “, ‘The main purpose of machine learning is to study and design the algorithms that can be used to produce the predicates from the given dataset, Besides these, the machine leaming includes the agents percepts for acting as well as to improve their future performance "A com experience ‘E’ with respect to some class of tasks performance at tasks in ‘T’ The following tasks must be learned by an agent, 1. To predict or decide the result state for an action. To keep record of relevant percepts, Machine learnin; efficiency of h 2 To know the values for each state(understand which state has high or low vale). 3, 4 8 Plays an important role in improving and understanding the uman learning. ~ Machine leaming is used to discover a new things not known to many human beings Methods ) Forms of Learning, BCA 2. Learning by taking advice 3. Learning in problem solving 4, Learning from examples: Induction 5. Learning by Decision trees. Rote learning is the basic learning activity. Rote learning is a memorization technique based on repetition. It is also called memorization because the knowledge, without any modification is, simply copied into the knowledge base. As computed values are stored, this technique can save a significant amount of time. Rote learning technique can also be used in complex learning systems provided sophisticated techniques are employed to use the stored values faster and there is a generalization to keep the number of stored information down to a manageable level. For example .Checkers-playing program The idea is that one will be able to quickly recall the meaning of the material the more one repeats it. Some of the alternatives to rote learning include meaningful learning, associative learning, and active learning. Ample, uses this technique to learn the board positions it evaluates in its look-ahead search. We mentioned one of the earliest game-playing programs, Samuel's checkers program [Samuel, 1963]. This program learned to play checkers well enough to beat its creator. It exploited two kinds of leaming/rote learning, which we look at now, and parameter (or coefficient) adjustment. Samuel's program used the minimax search procedure to explore checkers game trees. As is the case with all such programs, time constraints permitted it to search only a few levels in the tree. (The exact number varied depending on the situation) When it could search no deeper, it applied its static evaluation function to the board position and used that score to continue its search of the game tree. When it finished searching the tree and propagating the values backward, it had a score for the position represented by the root of the tree. It could then choose the best move and make it. But it also recorded the board position at the root of the tree and the backed up score that had just been computed for it. This situation is shown in Fig. (a). Now suppose that in a later game, the situation shown in Fig.(b) were to arise. Instead of using the static evaluation function to compute a score for position A, the stored value for A can be used. This creates ving searched an additional several ply since the stored value for A was acking up values from exactly such a search. the effect of computed by Tata PUBLICATIONS al[R] [a}[s|{r 10 Fig. Storing Backed-Up Values “ earning of this sort is very pane Itdoes not appear to involve any sophisticated sn soling capabilites But even it shows the need for some capabilities that will reining importantiin more complex learning systems. These capabilitesinclude Organized Storage of Information In order for it to be faster to use a stored value than it would be to recompute it, there must be a way to access the appropriate stored value quickly. InSamuel’s program, this was done ‘by indexing board positions byafew important characteristics, such as the number of pieces. Butas the complexity of the stored information increases, more sophisticated techniques are Necessary. 2 Generalization : The number of distinct objects that might potentially be stored can be very large. To keep the number of stored objects down to a manageable level, some kind of generalization is necessary. In Samuel's program, for example, the number of distinct objects that could be stored was equal to the number of different board positions that can arise in a game. Only a few simple forms of generalization were used in Samuel's program to cut down that number. All positions are stored as though White is to move. This cuts the number of stored positions in half. When possible, rotations along the diagonal are also combined. Again, though, as the complexity of the learning process increases, so too does the need for generalization. At this point, we have begun to see one way in which learning is similar to other kinds of problem solving Its success depends on a good organizational structure for its Tnowledge base, i ; das early as 1958 The idea of advice taking in AI based learning was proposed as ea es ade in creating such systems a. ‘ instructions writes a set of . grammer computer is @ is a teacher and the ¢ \y). Howeve few attempts were mi late T very few attempt ee Pn. systems providing a major impetus in this are b nee isa simple form of learning. Suppose 4 PO! tthe computer what to do, the programmer "Pi ea BCA i), the system will bein 9 POsition tg 4, i, rammed), i student. Once learned (i-¢ prog) : things. many sources! human ence nk a ee : ‘The advice may come oe inference than rote learning. ie nowledee “th . ires afer before stored in the knowledge base, M, oreo : tional fe knowledge should be considered. is conflicting wi Id ensure that the new knowledge is i g with the fe age 700 Fit OF tionaliser), for example, is a learnin, . knowledge. FOO (First Operational Operatto mpl i ea a Ige. It converts the advice wl ich is in the ie vhich i .d to learn the game of Hearts. which is use Bee aesthods into effective executable (LISP) Procedures ; t ly to use. This type of learning requ transformed into an opera reliability of the source of principles, problems, knowledge). Now this knowledge is read) There are two basic approaches to advice taking: Take high level, abstract advice and convertiit into rules thatcan guide performang elements of the system. Automate all aspects of advice taking : 2. Develop sophisticated tools such as knowledge base editors and debugging. Th 5 y Nes are used to aid an expert to translate his expertise into detailed rules. Here . expert is an integral part of the learning system. Such tools are i - . systems area of Al. SPETEaE I epee 1, LEARNING IN PROBLEM SOLVING . Learning to probl i iri k caring to pro chan, tite acquiring multiple forms of knowledge Problem solvingis viewed as a search of a ion of a problem. With this formalism arch of a state-s i PI : ; space formulati i i Operators are applied to states to transit from the initial State to th i te ie fe e goal state. lere are three basic me i i ‘ethods in which a System can learn from its own experiences. 4 Teaming by Parameter Adjustment é et by Macro Operators ee earning by Chunking ning by Parameter Adjustment Many programs 1 v SIAL INTELLIGENG E (he problem has an evalu: n fonction tha Coes 8 a polynomi al of eft He for™ such as ae Cgly bee + Cyl 5 of features and the c terms are weights ah the! terms & value: signiDB prog} tially. ‘dea of idea of parameter a caret x rams it is often difficult to decide on the exact val oe ini ee ie weight gp the basic i Start with some estimate of the correct weight settings. Modify the weight in the program on the basis of accumulated experiences Features that appear to be good predictors will have their weights i and bad ones will be decreased. heir weights increased seaming by Macro Operators se basic idea here is similar to Rote Learning: Avoid expensive recomputation Macro-operators can be used to group a whole series of actions into one. for example: Making dinner can be described a lay the table, cook dinner, serve dinner, We could treatlaying the tableas on action even though involves asequence of actions. The STRIPS problem-solving employed macro-operators in its learning phase. Consider a blocks world example in which ON(C,B) and ON(A,TABLE) are true STRIPS can achieve ON(A,B) in four steps: UNSTACK(C,B), PUTDOWN(C), PICKUP(A), STACK(A.B) MACROP with preconditions ON(C.B), STRIPS now builds a macro-operator p IN(C.TABLE) and the four steps #8 its ON(A,TABLE), postconditions ON(A,B), O body, MACROP can now be used in future operation. Butitis not very general. The above can be easily generalised wi Place of the blocks. How ith variables used in eneralisation is not always that easy le aming by Chunking Chunkin nd originates from sutational basis is in i ies to represent its a ae — Basic Outline of SOAR’s Method ; 1. SOAR solves problems it fires productions these are stored in long term BCA Memory, 2. Some firings turn out to be more useful than others. 3. When SOAR detects are useful sequence of firings, it creates chunks, 4 A chunk is essentially a large production that does the work of an entre sequency of smaller ones. Chunks may be generalised before storing, on is the process of assigning to a particular input, the name of which it belongs. The classes from which the classification procedure can choo described in a variety of ways, Classification is an important com solving tasks. In its simplest form, iti example of this is the question ” a class to Se can be mn task, An 2” But often classification happen, consider A to place B, and To use this rule successfully, an object as a wall. Without this, the system's matching routine must be able to identity system must be able to Tecognizi the rule can never be invoked. Then, to apply the rule, the a doorway, Before classific, done in a variety of . Is ‘ation can be done, the classes it will ‘use must be defined. This can be ct + Coty + cyt, +... —— GENCE the amp ‘ , Param yo! es Bs rainfall and Jocation of cold fronts. Different ‘itis Can be such i i, Ci ‘s oe parameters to predict sunny, cloudy, rainy, or sno ea = be writen to Y Weather, features that are relevant to the task domain. Defi '. Define each clas Sasa Je, if the task is weather prediction, eg get of f ysolate ms composed of those features. on Je, if the task isto identify animals, the body of each type of ani for ture, with various features representing such things as animal can be There are advantages and disadvantages to each ee length of stical approach taken by the first scheme presented ees more ral approach taken by the second. But the second is more fl ‘bl lexible asi al d feathers: eh pes. Testa sua est e ve excensible s of the way that classes are to be described, itis often difficult to construct, gor class definitions. This is particularly true in domains that are not ai that change rapidly. Thus the idea of producing a classification program gatcan evolveits own class definitions is appealing. This task of constructing class definitions scaled concept learning, oF induction. The techniques used for this task must, of course, depend on the way that classes (concepts) are described. If classes are described by scoring fancons, then concept Jearning can be done using the technique of coefficient adjustment. gegardles aaderstood or g Example — Winston (1975) Jinitions of concepts in this domain. a wedge (triangular block) suitably or an arch - two non-touching 4. Blocks World Learnin: The goal isto construct representation of the de! cincpts such a house - brick (rectangular block) with jaed ontop of it, tent - 2. wedges touching side by side, bids supporting a third wedge or brick, were learned. The idea of near miss objects — similar to actual instances was introduced. Input was Aine drawing ofa blocks world structure. Input processed (see VISION Sections later) 0 Pduce a semantic net representation of the structural description of the object ' pan pas-part oi E pject and semantic net Fig, : House ol oN » Links in network include left-of right-of, does-not-marry, supported-by,has-pars ang : Links oft-o : portant — two objects with a common touching gap, te uumed unless does-not-marry stated. BCA » The marry relation said to marry. Marrying is as There are three basic steps to the problem of concept formulation: 1, Select one know instance of the concept. Call this the concept definition. 2. Examine definitions of other known instance of the concept. Generalise the efiniton to include them, 3. Examine descriptions of near misses. Restrict the definition to exclude these. Both steps 2 and 3 rely on comparison and both similarities and differences need y be identified. 2. Version Spaces Structural concept learning systems are not without their problems. The bigges problem is that the teacher must guide the system through carefully chosen sequences of examples. In Winston’s program the order of the process is important since new links are added as and when now knowledge is gathered. The concept of version spaces aims is insensitive to order of the example presented. To do this instead of evolving a single concept description a set of possible descriptions are maintained. As new examples are Presented the set evolves as a process of new instances and near misses, We will assume that each slot in a version space description is made up of a set of Predicates that do not negate other predicates in the set positive literals.Indeed we can represent a description as a frame bases representation with several slots or indeed usea ‘more general representation. For the sake of simplifying the discussion we will keep 0 simple representations. If we keep to the above definition the Mitchell's candidate elimination algorithm is the best known algorithm. LEARNING BY DECISION TREES Quinlan in his ID3 system introduced the idea of decision trees, ID8 isa program that a baa jan, Muomatically from given positive and negative instances. Basically ex leaf of a decision tree asserts a positive or negat ticular gative concept. To classi ‘cular input we start atthe top and follow assertions do Sees EZ until we reach an answer. edible poisonous Fig. Edible Mushroom decision tree pulling Decision Trees D3 uses an iterative method. Simple trees preferred as more accurate classification is afforded. random choice of samples from training set chosen for inital assembly of tee — the window subset. 4 Other training examples used to test tree. 3, Ifall examples classified correctly stop. & Otherwise add a number of training examples to window and start again. ‘Adding New Nodes When assembling the tree we need to choose when to add a new node: | Some attributes will yield more information than others. Adding a new node might be useless in the overall classification process. ubsets whose members Sometimes attributes will separate training instances into s da leaf node assigned in @ common label. Here branching can be terminates an * the whole subset. Nivantages of De ecision Tree Monn version spaces when concept space is large. » of Decision Tree is declan to humans — a decision tree may find it hard to explain el, BOA Expert systems (8S) are one of the prominent research domains of AL, 1g intro the researchers at Stantord University, Computer Science Department. the "tex Stems are the computer applications developed to solve complex problems ina paiet domain, at the level of extraordinary human intelligence and expertise, Ula, Avexpertgystem isa computer program that simulates the judgementand beha -. Wid, pert knowledge and experience in a mt of a human or an orgat Patlane field. Typically, such as: ACCUMLaeg experience and a set of rules for applying the knowledge base to each Particular situa: that is described to the program, Sophisticated expert systems can be enhanceg With additions to the knowledge base or to the set of rules, tem contains a knowledge base containing Among the best-known expert systems have been those that lay chess and thay assist in medical diagnosis. An expert system is software that attempts to provide an answer to a problem, or clarify uncertainties where normally one or more human exper Would need to be consulted. Expert systemsare mostcommon ina specific problem domain, and is a traditional application and/or subfield of artificial intelligence (Al). A wide variety of methods can be used to simulate the performance of the expert however, common to most or all are: 1) The creation of a knowledge base which uses some knowledge representation structure to capture the knowledge of the Subject Matter Expert (SME); 2) A process of gathering that knowledge from the SME and codifying it according to the structure, which is called knowledge engineering; and 5) Once the system is developed, it is placed in the same real world problem solving situation as the human SME, {ypically as an aid to human workers or asa supplement to some information system. Expert systems may or may not have learning components Capabilities of the Expert System Below are some capabilities of an Expert System: 1. Advising: It is capable of advising the human being for the query of any domain from the particular ES. 2 Provide decision-mal ng capabilities: It provides the capability of decision mae in any domain, such as for making any financial decision, decisions in medical scien etc. Fos pavantages © 1 1 4 4, 5 3 s ear ‘aplications of Expert System 1 te a device: It is capable of demonstrating — GENCE a onstt@ ; ti ecifications, ow kK ij (0 use that product, etc any new prod i etc, ucts Such as its feat pn solVing: It has problem-solving capabili ilities, rr0! ini roblem: It i pxplaining ap is also capable of providit input ing a detailed de. snvcepreting u Seription of an esl FS designed for th: ie medical Ai ce piano? field is capable of ci it already contains ete disease inbuilt medica r oblem. he input: It i i inj p" is capable of interpreting the i input given by th 1 usi er, predicting results: It can be used for the i prediction of a result, ial osis: AN without using tools. f Expert System tems are highly reproducible. ‘These sys! ‘They can be used for risky places where the human pre: i Error possibilities are less if the KB contains satel ae nowledge. The performan’? of these systems remains steady as iti y Te ine ly as it is not affected by emotions, 3, The i i . ey provide a very high speed to respond to a particular que Limitations of Expert System . The response of the ex, rt syst i The por ystem may get wrong if the knowledge base contains ifferent scenarios Like a hut ing, i ‘l man being, it cannot produce a creative output for di maint tenance and development costs are very high. cult. «one ofthe bg limitations gning is much diffi pecific ES, which i nual updates. Knowledge acquisition for desi . each domain, we require a 5] t canno' i t learn from itself and hence requires ma In desieni i wa and manufacturing domain h ; ces such aS hae Fai eed or cee a sical devices aloes and automobiles. facturing PY othe use ae eknowled® ablishing ax a oF ne the releval an advisot an tems are primarily ‘used for P' ‘popular ES used for this dom BCA 3. In the finance domain F i In the finance industries, it is used to detect any type of possible fraud, . I sug activity, and advise bankers that if they should provide loans for busines, Picigy, rng i i 4. In the diagnosis and troubleshooting of devices In medical diagnosis, the ES system is used, and it was the first atea why, ty i systems were used. 5. . Planning and Scheduling The expert systems can also be used for planning and scheduling some Patio, tasks for achieving the goal of that task. Expert system is built around a knowledge base module. Expert system contains, formal representation of the information provided by the domain expert. This informa, may be in the form of problem-solving rules, procedures, or data intrinsic to the domain To incorporate these information into the system, it is necessary to make use of ‘one oy more knowledge representation methods. Some of these methods are described here Transferring knowledge from the human expert to a computer is often the mst difficult part of building an expert system, The knowledge acquired from the human expe must be encoded in such a way thatit remains a faithful representation of what the expert knows, and it can be manipulated by a computer. Three common methods of knowledge representation evolved over the years are 1. IF-THEN rules 2. Semantic networks 3. Frames 1. IE-THEN rules Human experts usually tend to think along : condition => action or Situation => conclusion Rules “if-then” are predominant form of encoding knowledge in expert system These are of the form : If ality... ,an Then bl bz, .! each ai is acondition or situation, and each bi is anactionor aconclusion. Poel usar caTions ARTIFI TCALINTELLigey ICE ai oe chemer knowledge is represented in te rms of obj jects and Telationships ee a & saver" denoted as ee ne denoted nodes of a graph. The relat ted a5 link between the corresponding t lationship betvee g two nodes. 1 two objects den t com yn form of semantic n qhe mos 7 jetworks uses nt I-A and HAS relationships betwe the links b ent objects tween nod 5 les to prose Je of gemantic Network Jow shows acar IS-A vehicle; a vehicle HAS whe wheels. he BS of relationship establish network inherit ork inheriting properties from the objects hi cts higher up. hes an inheril inheritance hy hiera in the network, ork, with the iis Kim : objec!s Jower down in the HAS Wheels Engine Battery | jeer called frames, situations, to handle sed into highly molar concepts, cl Knowledge consist" een concepts, a procedures 4 Frames Inthis techni bth pee knowledge is decomp a ae generalized record structures. nee a concepts, relationships betw ne ‘as well as attribute values: co ni a may be represented a8 4 separate frame. atti butes, scab bibgitee. oe relationships betwe' type numbers: The conte ntents of a slot may be of any data Proc res and so on. Ce ren concepts) strings functions or BCA iding the same kind of inp be linked to other frames, providing a ithe, The frames may be - tic network. that provided by a seman el hat pi based representation is ideally suited for obj al Peng, A frame-based r¢ techniques. i Example : Frame-based Representation of Knowledg t ‘Two frames, their slots and the slots filled with data type are shown, Wo fraines, Frame Pare cs Value [Value le Value Fi ESE Doors Vatve lvatue Is Value Working Memory Working memory refers to task-specific data for a problem. The contents of the working memory, changes with each problem situation. Consequently, it is the most dynamic component of an expert system, assuming that it is kept current. Every problem in a domain has some unique data associated with it Data may consist of the set of conditions leading to the problem, its parameters and 80 on. Data specific to the problem needs to be input by the user at the time of one, ‘means consulting the expert system. The Working memory is related to user inter! Fig, below shows how Working memory is Closely related to user interface of expert system. pee Lz Tava Pusticav@s ert system tool, or shell, is a software development environment containing : containit rt systems. ‘Associated with a shell is a prescribed method 10% f expel gand instantiating these components. Expertsystem faruilding applications by configurin; ally used for the purpose of allowing non-programmers to take advantage galls are basic ofihealready developed templates or shells and which have evolved because of the efforts {some pioneers in programming who have solved similar problems before. The core components of aN expert systems are the knowledge base and the reasoning engine. ‘An exP pasic components 0 ‘generic expert system ghell is shown below. _Aswecan see jn the figure above, the shell includes acquisition subsystem, an explanation subsystem and a user interface. new problem in any given. domain, we can find a shell which can provide .dge of an expert ‘or that problem, so all we need is the knowle: There are many commercial expert system shells available now shells a sient ange of problems. Taking help of PS" eae the = greatly reduces the cost and the time scopes npet system from the scratch. theinferenceengine,aknowledse When faced with a the rightsupport each one adequat® BCA Shell Components of a Generic Expert System 100 1. Knorwledge Base ion Subsystem 2 Knowledge Acq} Inference Engine Explanation Subsystem User Interface Knowledge Base Tt contains facts and heuristic knowledge. Developers try to use a Uniform, representation of knowledge as for as possible. There are many knowleg : Tepresentation schemes for expressing knowledge about the application domain ang some advance expert system shells use both frames (objects) and IF-THEN rules h PROLOG the knowledge is represented as logical statements. 4, 5. Knowledge Acquisition Subsystems The process of capturing and transformation of potentially useful information fora siven problem from any knowledge source (which may be a human expert) gg Programin the format required by that program is the job of a knowledge acquisition subsystem. So we can say that these subsystem to help experts build Knowledge bases. As an expert may not be a computer Uterate, so capturing information includes interviewing, preparing questionnaires ete, which is a very slow and time Consuming Process. So collecting knowledge needed to solve Problems and build the knowledge base has always been the biggest bottleneck in developing expert Some of the reasons behind the difficulty in collecting information are given below: 1. The facts and rules or principles of different domains cannoteasily be converted For example a teacher knows how to motivate the students but putting down on the paper, the exact reasons, causes and other factor affecting students i not easy as they vary with individual students, Different domains have their own terminolo; ry sau Ple-experts in particular domains w judgments, which information sources simplified, which is bas is very difficult, 4. Commonsense knowle capture, although syst hich information is important for specifi are reliable and how problems can te sed on personal experience. Capturing such knowledge found in humans Continues to be very difficult being made to capture it. : Tava PusucaToxs 4 TIF ICA — LINTE; Lig ENCE, The idea of automated “machine learning” ee ae capturing has also been 9: goal is that, a computi the important research nein mom, goal is tha a computing system or machine could atea fo SMM fy problems like the way human do it. tld be enabled te fo HOw, ise : 0 lean. Inference Engine team in order . ‘An inference engine is w sed to perform reasoni euereticg 0 yn is extracted from an expert and most aa is specific to the problem being solved, Bron, Buan ex form of a set of IF-THEN rules, The case specific an oo edge wnat the user and also partial conclusions Gan wian includes the data prong : it i . Ovi this data, In a normal forward chaining rule-based cit certainty facto) bases the elements in the working memory S¥stom the case specie cific data is 8 with both the expert knowled, e Different strategies for using domain-specific knowled, eee Je Fe ce characteristics of programs, and also on the ; ; way in whi finds or searches a solution among possible Seana in which a program representations schemes are sed under varity of rasa S. Most knowledge is going on in this area. ning methods and research ige have great effects on the Explanation Subsystems ‘An explanation subsystem allows the program to explain its reasoning to th ‘The explanation can range from how the final or intermediate tafitons ee aa at to justifying the need for additional data. fe artived Explanation subsystems are important from the following points of view : .dge: There must be some for the satisfaction of knowledge a. Proper use of knowle time of development engineers that the knowledge is applied properly evenat the of a prototype. b. Correctness of conclusions: conclusions produced by the system are correct. c. Execution trace: In order to judge that the knowledge elicitation is proceeding ete trace of program execution is required. smoothly and successfully, a comp! haviour: For proper maintenance and debugging, behaviour is necessary for the programmers: i stems are nece' oach: Explanation subsystems @ I fea js suitable to the particular domain se the reasoning Ce ue of control P Janation solution. Ex for, iuced by @ User's need to satisfy themselves that the the knowledge of program reasoning appr asoning technique apP deals with the iss it jepend on hot nee ean’ ; related to evaluation 4s y formed while reasoning" its act g the trace of its im 5. User interface BCA be decided that whether or not a system is producing the right answer ¢, reasons, Without a good explanation subsystem, an expert will be unable 4, "th the system’s performance or will be unable to find ways to improve jt, Mpe Itis used to communicate with the user. The user interface is generally not os tof the expert system technology, and was not given much attention in the However, it is now widely accepted that the user interface can make a ght cal difference in the utility of a system regardless of the systems performance, Types of Expert System Sell 1. 2. 3. There are some important types of Expert System Shell are as follows; C Language Integrated Production System (CLIPS) C Language Integrated Production System (CLIPS) is an expert system shell fo, building expert systems. It is the most widely used expert system shell. Itis written in C programming language. Initially, LISP was the single tool available with NAS, for all expert system software . This tool was having three major drawbacks lie. Iow availability of LISP on a wide variety of conventional computers, high cost tp set-up system using LISP and poor integration with other languages Essential MYCIN (EMYCIN) EMYCIN (Essential MYCIN) is an expert system shell derived from MYCIN - an expert system developed for diagnosis of infectious diseases. The MYCIN was one of the best known expert system developed by Dr. Edward Shortliffe in 1974 for medical diagnosis . As MYCIN was domain dependent system, it could not be used for the development of other domain specific expert system. At that time the knowledge engineer were searching for a system that could be used for building expertsystem for multiple domains. Thanks to Dr. Crawford, who in 1987 had derived an independent system from MYCIN that is domain independent. It does not contain domain specific knowledge base. It means its knowledge base is blank intelligentsyst™ in cts as an interpreter for Jess language. It was developed : the Jess engin® +. It contains * “Jess can beus Wrledge youse?h in also be called / executed by Ji . So, we can say it’s a superset o E-rara Pusat iyayn Ine ‘The Kxper tou Experts Faye Ita rulesbased expert ystom gh the Mout elfciont and effective way to automate the dclivery of atv he ular ety to lea ond erential the ga asa toexplnin why they made a decoion, Others une ae aM expert yn ning ett would. rules ina true “Neus oul Use. calling andl using rules dynamically a9 they are needled ig backw, ole” way, Bxays Inference Ligne provides both options, ant they can be mf ining The ral . floxibility, ied OF maximum Vidwan An expert syatem Shell - Vidwan has been developed at the Nation, st) 1 1998. 1 are (NCSI), Mumbai (India) in 1993, Its a web-based expert g ster toereate rule-based expert system, As itis. rule-hage oe Sell tha ws toencode domain knowledgein the form of THIEN ofthis shell applies backwardchaining inference mechanta y of information, The explanation component o the shen Provides fata te Piece an specifi query inthe form of why and how. The lates of Vidwan compatible with UNIX operating system, idwan is also Morits of Expert System Shell There is no need to build shell again and aj gain for new expert system. Hence, the time required in building user interface ani d inference engine is not needed, Development of an expert system becomes faster, cheaper and more efficient, Itdoesn’t require more experience Programmer because there is no need to develop the system from scratch, Domerits of Expert System Shell L hv and how I the base is weak then developed product may not be reliable. t means i the shell isnot matching the system requirement then the developed system will not represent expertise of a domain expert. Some expert systems Tequire different look and fill, security level and user guide, Thisis not possible ina single expert system shell, It needs customization depending on expert system requirement. 3 Most expert syste *xplanation facilities that allow the user to ask questions ee » advice on car propy al to give ad blony Uyample: nod 10 f 4, Dialog w System Do om desl ith an expert aystom ie nea the car start ? User No. ne » turn over System Does the engine turn 0 User Yes. System Do you smell gas ? User Yes inule: ry again, System Recommendation ~ Wait'S minutes and try aga User Why ? System I used the rule: If not start, and engine_turn_over, and smell_gas Then recommendation is ‘Wait 5 minutes and try again’ User Note: The rule gives the correct advice for a flooded car, and knows the questions to be ask to determine if the car is flooded, but it does not contain the knowledge of yy a flooded car is and why waiting will help. Types of Explanation There are four types of explanations commonly used in expert systems. Rule trace reports on the progress of a consultation; 1. Explanation of how the system reached to the given conclusion; 2. Explanation of why the system did not give any conclusion. 3. Explanation of why the system is asking a question, Knowledge acqui fora knowled, ‘0 describe the initial tas] and interviewing doma frame based ontologies - Expert systems we technology to real world: “laboratories. worked with di © that could automate had mostly be re and other hi asks such as me envused to automate highly data in Technologies such as inference more complex problems, BHIY skied ey ‘lical agnosis, nlensive engines allowed dey, 8 0 dey ; lo comple Unt ai A 3 (sks but oy Clopers for easoni he first ing © flat time ty As expert systems sealed ay applications it was soon realize one of if not the most critical ta acquisition process became P from, demonstration Protot i 8 l0 induste; Acquisition of domain, expert bee enath knowledge engineering process, This Was An intense aren of research on ite own, S knowledge Issuos In Knowledge Acquisition id that the k in’ the ‘The important issu in knowledge acquisition aro: 1, knowledge is in the head of experts, 2, Experls have vast amounts of knowledge 3, Experts have a lot of tacit knowk ledge-They do not seTacit knowledge is hard ( OW all that they fy impossible) to describe 7 know and 4, Experts are very busy and valu able people 5 One expert does not know everything 6, Knowledge has a “shelf life” Techniques for Knowledge Acquisition Protocol analysis tecliniques : Used Wilh transcripts of interviews or text-based information to identify basic knowledge objects within a protocol, such as goals, decisions, relationshi tributes. These act asa bridge between the use f Protocol-based techniques and knowledge modeling techniques Hiera hy-generation ‘echniques : Involve creation, reviewing and modification of hiera hical knowledge. Hiera hy-generation techniques, such as laddering, are eet '9 build taxonomies or other hiera hical structures such as goal tees and decision . The Ladders are Of various forms like concept ladder, attribute ladder, “position ladders, Mattixbased techni ‘abla, indicat iques : Involve the construction and filling-in a 2-D ae ve "such things, as may be, for example, between concepts and prope "Paco 3 BCA ~ (attributes and values) or between problems and solutions or between tasks ang re80ule9, tc. The cletienta within the matrix can contain: symbols (ticks, Cross, ANestion marks ), colors, numbers, text. Sorting techniques : Used for capturing the way people contpare and order ¢ it may reveal knowledge about classe , properties and priorities. Limited-information and constrained-processing tasks : Techniques thateither limit ime and/or information available to the expert when performing tasks, Fer example, a twenty-questions technique provides an efficient way of accessing the oncepts, niques : Include generation and use of concept maps, state ent diagrams and process maps. These are particu larly the “what, how, when, who and why” of tasks and events, IMPORTANT QUESTIONS What is learing ? Explain about Rote tearing, Leaming by taking Advice, learning in problem ‘solving and Learning by Decision ‘trees, What is Expert System? Explain about Representing and Using Domain Knowledge. Write about Expert systems shells and Explanation, Explain about Knowledge Acquisition

You might also like