Presentation is loading. Please wait.

Presentation is loading. Please wait.

 Rosina Weber Case-based reasoning INFO 612 Dr. R. Weber.

Similar presentations


Presentation on theme: " Rosina Weber Case-based reasoning INFO 612 Dr. R. Weber."— Presentation transcript:

1  Rosina Weber Case-based reasoning INFO 612 Dr. R. Weber

2  Rosina Weber Where does it come from? Pursuit of representation of memory: Schank’s Dynamic Memory (1982) concept (i.e. scripts, MOPS, links ) role of understanding in solving problems similarity heuristic the reminding of a past episode that is similar to a current one so that one can apply a strategy/solution that has worked in a similar episode

3  Rosina Weber CBR assumptions hypotheses similar problems have similar solutions problems recur (Leake, 1996)

4  Rosina Weber Definitions From Riesbeck & Schank (1989), "A case-based reasoner solves new problems by adapting solutions that were used to solve old problems". Case-Based Reasoning systems mimic the human act of reminding a previous episode to solve a given problem due to the recognition of their affinities. Case-based reasoning is a methodology that reuses previous episodes to approach new situations. When faced with a new situation, the goal is to retrieve a similar previous one and reuse its strategy.

5  Rosina Weber CBR and AI tasks  Mundane  prediction-advice  composition  understanding  reading  planning  walking  uncertainty  creativity Both  interpretation  classification  categorization  discovery  control  monitoring  learning  planning  analysis  explanation Expert  diagnosis- troubleshooting  prescription  configuration  design  scheduling  retrieval  mediation  argumentation  recommendation

6  Rosina Weber CBR applications deployed emerging research

7  Rosina Weber Deployed

8  Rosina Weber CBR applications CCBR conversational CBR e gain Examples from Lucas Arts e gain Examples from Lucas Arts

9  Rosina Weber PROFIT valuates residential properties to evaluate mortgage packages for a division of GE Mortgages. Values of a property change with market conditions, so estimates have to be updated constantly according to real estate transactions, which validate the estimations. CARMA is designed to provide expert advice on handling rangeland grasshopper infestations. CARMA has reused its expertise combined with model-based methods to devise policies on pest management and the development of industry strategies.

10  Rosina Weber General Motors has developed an organizational CBR system to support the goals of dimensional management, an area in the manufacturing of mechanical structures (e.g., vehicle bodies) that enforces quality control by reducing manufacturing variations that occur in fractions of millimeters. Western Air is an Australian distributor of heat and air conditioning systems; they have chosen to use a web-based CBR application [20] to guarantee a competitive advantage that also poses an entry barrier to competition. They guarantee the precision of the specifications of each new system and the accuracy of the quotes by relying in knowledge captured in previous installations.

11  Rosina Weber PTV combines case-based (content- based) personalization with collaborative filtering to recommend shows to watch on digital television. NEC has developed SignFinder, which is a system that detects variations in the case bases generated automatically from customer calls. When they detect variations on the content of typical customers requests, they can discover knowledge about defects on their products faster than with any other method.

12 Compaq SMART Diagnosis and repair; customer support help desks Acorn, Walden Uses Inference ’ s tool; can be used by up to 60 users at a time; shows that library engineering is necessary ALFAPredict power demandJabourSame result but faster than human experts CLAVIERDesign and evaluation of autoclave loading Barletta & Hennessy Interacts planning and scheduling SQUADSoftware quality control advisor Kitano20,000 cases in 1993 HVAC systemTests and diagnosis of faults in A/C systems Watson, 2000Diagnosis and solutions to HVAC maintenance Operated by salespersons Western Australia FormToolCBR in color matchingCheethamGE CRD Savings of 2.25 million per year in productivity and cost reduction PTV (personalized TV listings) Each user receives a daily personalized TV listing specially compiled to suit each user’s individual preferences Cotter & Smyth Cbr and collaborative filtering CF makes a recommendation to a person because his or her profile is similar to other people who have chosen the recommended item.

13  Rosina Weber Emerging

14  Rosina Weber Dublet recommends apartments for rental in Dublin, Ireland, based on a description of the user’s preferences. It employs information extraction from the web (of apartments for rent) to create cases dynamically and retrieves units that match the user’s preference. Dublet performs knowledge synthesis (creation) and extends the power of knowledge distribution of the CBR system by being operational in cell phones.

15  Rosina Weber The Auguste Project CBR is used to decide whether a patient benefits from a drug and RBR decides which drug to choose Marling 2001Planning ongoing care for AD (Alzheimer) cases based on strategies that worked better in past cases DUBLETRecommends rental properties from different online sources Hurley, Wilson 2001 Is used on the web and in mobile phones Employs Information Extraction tools to gather info from the web- returns properties ranked according to similarity HICAP

16  Rosina Weber Research

17 nametaskauthorobs. ABBYRomantic advisor; retrieves a similar history DomeshekSocial context ARCHIE ARCHIE 2 Architecture design of office buildings Goel, Kolodner and Domschek CADETDesign of mechanical components Sycara, Navinchandra Abstract indexing allowed innovative design CASEY Diagnosis cause and prescribes solution to heart problems Kotonmodel-based CHEFDesign of recipes to meet different simultaneous goals Hammond case-based planning: Memory started with 20 recipes and learned from user feedback COACHPlanning soccer gamesCollinsDebugging and fixing bad strategies; memory keeps strategies and the type of problem HYPOInterpretation and argumentation Rissland & Ashley Retrieves similar cases to create a point, a response, and a rebuttal using hypotheticals (Ashley, 1990) JUDGE Defines sentences of delinquent crimes based on the chances of repeating the crime and its severity BainIn case of not having a sufficient similar case, the system uses heuristics to determine the sentence JULIA planning mealsHinrichsPlausible reasoning and design

18  Rosina Weber nametaskauthorobs. CYRUSstored and retrieved events in the life of Cyrus Vance when he was secretary of state Kolodner first implementation of MOPs reconstructive dynamic memory AQUA story understanding, explanation on terrorism Ram reads newspaper stories and asks questions, learning through incremental revision of knowledge; case-based explanation CASCADEassistance on recovering from crashes in VMS OS Simoudis & Miller (first) help desks; emphasis on efficient retrieval when first descriptions are not rich ASK user directed exploration of stories and guidelines describing a task or domain Ferguson, Bareiss, Schank ASK Tom trust bank consulting; ASK Michael industrial dvlpmnt CELIA automated diagnosis and interactive learning; predicts an expert’s action and relate steps Redmond acquiring cases, learning indexes, combines cbr and other methods Mostly from Kolodner 1993

19 nametaskauthorobs. MEDIATO R Mediates conflicts by performing planning SimpsonKeeps in memory failed solutions and tries to avoid same failures in new solutions PERSUADE R Mediation of union negotiations; proposes solutions with arguments SycaraConsiders part’s goals and considers recent accepted solutions AMADEUSsuggests how to write papers Aluisio, 1995 PLEXUSPlanning daily tasksAltermanAdapts the experience of riding the SF metro to reuse in NY PRODIGYPlanning and learningVeloso, Carbonell Demonstrated in a variety of domains PROTOSHeuristic classification for diagnosis Bareiss, Porter, Murray, Weir, Holte Automatic knowledge acquisition; good for weak theory domains SWALEGenerates explanation of anomalous events in news stories Schank, Kass, Leake, Owens Searches for similar explanations for death and destruction such as the murdered spouse that was killed because of the insurance money just like the horse (SWALE) that was killed by its owner for the same reason Mostly from Kolodner 1993

20 nametaskauthorobs. CATOTutoring systemAleven/Ashle y Teaching law students to create argument HICAPCase-based planningMunoz Avila 1999 Combines case-based planning with methods in planning NEO’s PRUDENTIAJurisprudence research; textual CBR Weber, 1998Case retrieval Recent applications Springer series on CBR Research and Development

21  Rosina Weber CBR systems types interpretive: –past cases are used as references to categorize and classify new cases –interpretation, diagnosis problem-solving: –past cases are used to provide a solution to be applied to new cases –design, planning, explanation, lessons learned

22  Rosina Weber case base case representation CBR methodology Task?

23  Rosina Weber case base situation assessment CBR methodology

24  Rosina Weber case base RETRIEVE REUSE REVISE RETAIN CBR methodology

25  Rosina Weber CBR development life-cycle Knowledge acquisition CBR design and implementation – Situation assessment – Retrieve – Revise – Review – Retain Validate Maintain

26  Rosina Weber Knowledge acquisition From humans From data From databases From text: textual CBR Machine learning (e.g., data mining to learn cases)

27  Rosina Weber CBR design and implementation Situation assessment Retrieve Revise Review Retain (Validation) Maintenance design

28  Rosina Weber case base Design decisions in CBR systems (i) Which are the cases? What is the task? How will the case base be organized? How will the cases be represented? Which will be the indexing vocabulary? What is the task? How will the case base be organized?

29  Rosina Weber How will new cases be input? How to perform retrieval? Identify features Initially match (similarity assessment) Search Select Retrieval input problem initial solutions Design decisions in CBR systems (ii)

30  Rosina Weber How to implement reuse? From Select or with a combination? How to display the proposed solution? solution Reuse proposed initial solutions Design decisions in CBR systems (iii)

31  Rosina Weber Is the proposed solution good? How to determine and find what to adapt? Where is adaptation knowledge? solution Revise proposed confirmed solution case repair case adaptation Design decisions in CBR systems (iv)

32  Rosina Weber Is it the type of task that it is worth learning? Index new case before retain. Retain. Retain confirmed solution case base Design decisions in CBR systems (ii)

33  Rosina Weber Knowledge in case-based reasoning systems by Richter, M. M., “The Knowledge Contained in Similarity Measures: Some remarks on the invited talk given at ICCBR'95 in Sesimbra, Portugal, October 25, 1995”. Online: http://www.cbr- web.org/documents/Richtericc br95remarks.html

34  Rosina Weber Validation refers to establishing the effectiveness of a system in light of its intended purposes Verification indicates how correct a given system can solve its proposed tasks (Watson) Retrieval accuracy is indicated by the result given by the system when the target case is part of the case collection. validation & verification (i)

35  Rosina Weber validation & verification (ii) Retrieval consistency: the same retrieval when executed the second time must retrieve exactly the same cases (e.g., with the same similarity if k- NN is used) Case Duplication: when two distinct cases receive the same value for similarity in relation to a given target case. When the same value is attributed to different cases the user or the system has to decide which one to use by evaluating the value for each attribute. The same measure of similarity does not mean the cases necessarily teach the same lessons.

36  Rosina Weber Case Coverage is checked for the evenly distribution of cases when they are manipulated and not actual experiences that are collected as they happen. Efficiency verification : comparison to alternative methods, empirical evaluations Retrieval time Retrieval sorting Case base consistency can be indicated by retrievals resulting cases with gradual values of similarity. A retrieval that no case has a high value of similarity or too many cases have the same value suggests inconsistency in the case base validation & verification (iii)

37  Rosina Weber maintenance if the reasoner learns, the maintenance is more elaborate statistics of case usage perform validation tests continuously special issue on case-based maintenance Neural networks and other soft computing methods have been proposed methods for distributed case bases

38  Rosina Weber Advantages of CBR systems (i) Knowledge acquisition and representation: There is no need to explicit acquire and represent all the knowledge the system can use. Common sense: knowledge that would have to be represented explicitly is implicitly stated in cases. Not easily formalizable tasks: such as in some medical domains, prototypical descriptions represent more easily a body of knowledge. Creativity - Case solutions can be combined into new ones and cases can also be used in a different level of abstraction providing innovative solutions.

39  Rosina Weber CBR systems can avoid mistakes Learning - can be done without human interference; CBR systems can become robust and provide better solutions. User’s feedback is easily incorporated in the revise phase. Degradation -CBR systems can recognize when no answer exists to a problem by simply defining a threshold from which a solution is no longer acceptable. In decomposable problem domains, a solution can be created from the combination of partial solutions. (shared with ES and other AI) Permanence - CBR do not forget. Breadth - One CBR system can entail knowledge learned from an unlimited number of human experts. Reproducibility - Many copies of a CBR system. Advantages of CBR systems (ii)

40  Rosina Weber books, resources Leake, D. (1996). Case-Based Reasoning: Experiences, Lessons, and Future Directions. AAAI Press/The MIT Press, Menlo Park, California, 1996. Kolodner, J. (1993). Case-Based Reasoning. Morgan Kaufmann, Los Altos, CA. Watson, Ian (1997). Applying Case-Based Reasoning: techniques for enterprise systems. Morgan Kaufmann Publishers, Inc. San Francisco, California. Schank, R. (1982). Dynamic Memory: A theory of learning in computers and people. New York, Cambridge University Press. Schank, R., Kass, A. and Riesbeck, C. (1994). Inside case-based explanation. Lawrence Erlbaum Assoc., Hillsdale, N.J. Ashley, Kevin D. (1990). Modeling Legal Argument: reasoning with cases and hypotheticals. A Bradford book. The MIT Press, Cambridge, Massachussetts. Lecture Notes in Computer Science, CBR research and development, Springer Lecture Notes in Computer Science, Advances in CBR, Springer ai-cbr.org

41  Rosina Weber Building (shells), using, maintaining Shells/tools –http://www.cbr-web.org/CBR-Web/?info=tools&menu=pthttp://www.cbr-web.org/CBR-Web/?info=tools&menu=pt –Esteem examples, NISTP CBR Shell examples Using –Laypeople, experts Maintaining –Automatically learning new cases Cases are real or created –Manually adding new cases

42  Rosina Weber

43 Further reading Riesbeck & Schank (1989) Inside case- based reasoning Kolodner (1993) Case-based reasoning Aamodt & Plaza (1994) AICom paper Leake (1996) Leake, David. (1996). Case-Based Reasoning: Experiences, Lessons, and Future Directions. Watson (1997) Applying Case-Based Reasoning: techniques for enterprise systems.

44  Rosina Weber Use the mode Do the first 4 cases have the same similarity? Yes* No Do the first 5 cases have the same similarity? Is there conflict? Do the first 3 cases have the same similarity? Yes** No *No conflict possible; ** Conflict possible Yes No Use result from the 5th.Use the mode No Do the first 2 cases have the same similarity? Yes* Use the mode No Use the result from the most similar Yes** Is there conflict? Yes No Use result from the 3rdUse their result


Download ppt " Rosina Weber Case-based reasoning INFO 612 Dr. R. Weber."

Similar presentations


Ads by Google