6.871 Lecture 9 1Frame-Based Systems6.871 Lecture 96.871 - Lecture 9 2Outline Minsky’s original motivations, observations Details and use In the spirit: PIP and Internist-1 Not in the spirit: FRL Frames summary Comparison of KR technologies6.871 - Lecture 9 5A KR Should Tell You What to attend to:“A Frame …[represents] a stereotyped situation…” What inferences are recommended:“When one encounters a new situation …, one selects from memory a structure called a frame… a remembered framework to be adapted to fit reality by changing details as necessary.”Minsky “A Framework for Knowledge Representation”6.871 - Lecture 9 19Motivations and Observations A model of human cognition; the structure of knowledge memory; “common sense” reasoning Explain why understanding is …–fast– anticipatory– persistent over changes in perspective– tenacious: “Colorless green ideas sleep furiously.” Meaning is poorly approximated by dictionary defns. Memory is full of prototypical situations, richly interconnected.6.871 - Lecture 9 21Use Frames are a useful representation when the task is to understand [or explain, react to] a new situation.“When one encounters a new situation …, one selects from memory a structure called a frame… a remembered framework to be adapted to fit reality by changing details as necessary.”6.871 - Lecture 9 22Details Frames are networks– Top levels fixed– Lower levels hold specific instances of data– Terminals holding data have easily displaced defaults Inferencing is matching of data to prototype– Subjective, approximate Optional (in the original conception):– Hierarchy of frames, inheritance– Daemons: procedures triggered when needed6.871 Lecture 9 26.871 - Lecture 9 23ExampleBirthday Party6.871 - Lecture 9 25In The Spirit: PIP Motivated by data on clinical cognition:– Quick focus on little data– Not easily refocused– Ask discriminating questions– Answer is an ordered list of matches Wanted expert level performance6.871 - Lecture 9 26In The Spirit: PIPNephroticSyndromeIS-A ClinicalStateFinding Low Serum AlbuminFinding Heavy ProteinuriaFinding …MustNotHave Proteinuria AbsentSufficient Pedal edema and proteinuria > 5gm/dayMayBeCausedBy Acute GlomerulonephritisMayBeCompBy HypovolemiaScoringEdema: Massive, symmetrical: 1.0Not massive, symm. 0.5Asymmetrical -0.5… 70 Disease frames, 500 findings Variety of interconnections: MustNotHave, ComplicatedBy…6.871 - Lecture 9 27PIP’s Machinery Hypothesis generation via data-driven triggering– Frame moves into short term memory– “Nearby” frames become semi-active Hypothesis testing via calibrating match of data & frame– Match of frame and data» Sufficiency, exclusionary rules»Scoring– Ability to explain the findings Additional data gathering to fill terminals– Asks questions6.871 - Lecture 9 28 Doctors move from more general to more specific disorders– Need hierarchy of framesALCOHOLIC HEPATITISAKO HepatitisFindingsAge 16-25 0 1Age 26-55 0 3Age >55 0 2Alcohol History 2 4Causes Hepatatic Encephalopathy 2 2Hierarchy, rooted on organ systems The numbers: evoking strength and frequency 500 disease frames, 3500 findingsIn the Spirit: Internist-16.871 - Lecture 9 29Internist-1: Reasoning Begin with lots of data Evoking strength determines active hypotheses– increased/decreased for present/absent findings Matching controlled by “undershoot” and “overshoot” Reasoning strategies– pursue, rule out, discriminate6.871 Lecture 9 36.871 - Lecture 9 30Not in the Spirit: FRL Task: a scheduler constraint propagation + common sense Hierarchical frames; viewed as “property lists” Wide variety of explicit slot types, e.g.:– Comments (source of value)– Defaults– Value– Constraints on values Attached procedures– IfNeeded, IfAdded, IfRemoved Looks like? 6.871 - Lecture 9 31FRLMEETINGAKO VALUE ActivityWHO REQUIRE EXIST x Chairman(x)WHENRA-GROUP-MEETINGAKO VALUE MEETINGWHERE DEFAULT ConferenceRoom1WHEN DEFAULT FridayPREFER WeekdayACTIVITYAKO VALUE THINGWHEN IfAdded AddToCalendar6.871 - Lecture 9 32Not in the Spirit: FRL Where is the theory of intelligent reasoning? Where are the “glasses”? Instead of knowledge representation we have…? A common mistake: focus on mechanisminstead of intent.6.871 - Lecture 9 33Frames Summary Inspired by human understanding and reasoning Prototypes and matching as key concepts Representations evolve: Originally a model of human memory and cognition, now at times used more mechanistically 6.871 - Lecture 9 34Representation and reasoning usingLogic:bird(x) can-fly(x)Rules: If class of animal is bird then animal can fly (.9)SI-Nets:Frames:BirdClass AnimalLoco FlyComparing the TechnologiesAnimalFlyLoco6.871 - Lecture 9 35Comparing the TechnologiesGranularity of unit of meaning Logic–Axioms Rules– Centered around heuristic association– Individual inference step SI-Nets– Organized around “nouns”– Necessary and sufficient conditions Frames– Organized around prototypes– Meaning spread throughout the network.6.871 Lecture 9 46.871 - Lecture 9 36Comparing the TechnologiesReasoning Logic– Formal deduction– Results precisely determined Rules– Chains of heuristic associations– Uncertainties combined SI-Nets– Logic-based subsumption algorithm– Formal method and result Frames– Heuristic matching of instances to prototypes– Ranked by