Learning Agents Center George Mason University .


50 views
Uploaded on:
Description
. Critical thinking Method: Task Reduction. Learnable Knowledge Representation: Plausible Version Spaces. Multistrategy Learning amid Problem Solving . Review. . Operators Development Experiments. Showing and Learning Demo. Affirmations. Research Problem, Approach, and Application. Future Directions: Life-long Continuous Learning.
Transcripts
Slide 1

Pupil Reasoning and Learning Agents Gheorghe Tecuci with Mihai Boicu, Dorin Marcu, Bogdan Stanescu, Cristina Boicu, Marcel Barbulescu Learning Agents Center George Mason University Symposium on Reasoning and Learning in Cognitive Systems Stanford, CA, 20-21 May 2004

Slide 2

Overview Research Problem, Approach, and Application Problem Solving Method: Task Reduction Learnable Knowledge Representation: Plausible Version Spaces Multistrategy Learning amid Problem Solving Agent Development Experiments Future Directions: Life-long Continuous Learning Teaching and Learning Demo Acknowledgments

Slide 3

Research Problem and Approach Research Problem: Elaborate a hypothesis, strategy and group of apparatuses for the advancement of information base specialists by topic specialists, with restricted help from learning engineers. Approach: Develop a learning operator that can be instructed specifically by a topic master while tackling issues in participation. The master educates the operator to perform different errands in a way that takes after how the master would instruct a man. The specialist gains from the master, assembling, checking and enhancing its learning base 1. Blended activity critical thinking 2. Educating and learning 3. Multistrategy learning Problem Solving Ontology + Rules Interface Learning

Slide 4

Sample Domain: Center of Gravity Analysis The focal point of gravity of a substance (state, union, coalition, or gathering) is the establishment of ability, the center point of all power and development, whereupon everything depends, the point against which every one of the energies ought to be coordinated. Carl Von Clausewitz, On War, 1832. The focal point of gravity of a substance is its essential wellspring of good or physical quality, power or resistance. Joe Strange, Centers of Gravity & Critical Vulnerabilities, 1996. On the off chance that a warrior disposes of or impacts the adversary\'s vital focus of gravity, then the foe will lose control of its energy and assets and will in the long run tumble to overcome. On the off chance that the warrior neglects to satisfactorily secure his own vital focus of gravity, he welcomes debacle. Giles and Galvin, USAWC 1996.

Slide 5

First computational way to deal with COG investigation Approach to focus of gravity examination in view of the ideas of basic capacities , basic prerequisites and basic vulnerabilities , which have been as of late received into the joint military principle. Application to current war situations (e.g. War on fear 2003, Iraq 2003) with state and non-state on-screen characters (e.g. Al Qaeda). Recognize COG hopefuls Test COG competitors Identify potential essential wellsprings of good or physical quality, power and resistance from: Test each distinguished COG contender to figure out if it has all the vital basic capacities: Which are the basic abilities? Are the basic necessities of these capacities fulfilled? If not, dispose of the hopeful. In the event that yes, do these capacities have any helplessness? Government Military People Economy Alliances Etc.

Slide 6

Overview Research Problem, Approach, and Application Problem Solving Method: Task Reduction Learnable Knowledge Representation: Plausible Version Spaces Multistrategy Learning amid Problem Solving Agent Development Experiments Future Directions: Life-long Continuous Learning Teaching and Learning Demo Acknowledgments

Slide 7

Problem Solving: Task Reduction T 1 An intricate critical thinking undertaking is performed by: progressively lessening it to more straightforward assignments; finding the arrangements of the least complex errands; progressively making these arrangements until the answer for the underlying assignment is gotten. S 1 Q 1 … S 11 A 1n A 11 S 1n T 1n T 11a S 11a T 11b S 11b … Q 11b S 11b … S 11bm S 11b1 A 11bm A 11b1 … T 11b1 T 11bm Let T1 be the critical thinking assignment to be performed. Finding an answer is an iterative procedure where, at each progression, we think of some as applicable data that leads us to decrease the present undertaking to a less complex assignment or to a few more straightforward errands. The question Q related with the present assignment distinguishes the sort of data to be considered. The appropriate response A distinguishes that snippet of data and leads us to the diminishment of the present errand.

Slide 8

COG Analysis: World War II at the season of Sicily 1943 We have to Identify and test a key COG possibility for Sicily_1943 Which is an opposing_force in the Sicily_1943 situation? Allied_Forces_1943 Therefore we have to Identify and test a key COG possibility for Allied_Forces_1943 Is Allied_Forces_1943 a single_member_force or a multi_member_force? Allied_Forces_1943 is a multi_member_force Therefore we have to Identify and test a key COG contender for Allied_Forces_1943 which is a multi_member_force What sort of vital COG competitor would it be advisable for me to consider for this multi_member_force? I consider a hopeful comparing to an individual from the multi_member_force Therefore we have to Identify and test a key COG competitor relating to an individual from the Allied_Forces_1943 Which is an individual from Allied_Forces_1943? US_1943 Therefore we have to Identify and test a vital COG contender for US_1943

Slide 9

Overview Research Problem, Approach, and Application Problem Solving Method: Task Reduction Learnable Knowledge Representation: Plausible Version Spaces Multistrategy Learning amid Problem Solving Agent Development Experiments Future Directions: Life-long Continuous Learning Teaching and Learning Demo Acknowledgments

Slide 10

Knowledge Base: Object Ontology + Rules Object Ontology A various leveled portrayal of the items and sorts of articles. A various leveled portrayal of the sorts of components.

Slide 11

Knowledge Base: Object Ontology + Rules We have to Identify and test a key COG competitor comparing to an individual from the Allied_Forces_1943 Which is an individual from Allied_Forces_1943? Case OF REASONING STEP US_1943 Therefore we have to Identify and test a key COG contender for US_1943 LEARNED RULE IF Identify and test a vital COG hopeful comparing to an individual from a compel The drive is ?O1 IF Identify and test a key COG competitor relating to an individual from the ?O1 Plausible Upper Bound Condition ?O1 is multi_member_force has_as_member ?O2 is force Question Which is an individual from ?O1 ? Reply ?O2 Plausible Lower Bound Condition ?O1 is equal_partners_multi_state_alliance has_as_member ?O2 ?O2 is single_state_force THEN Identify and test a vital COG contender for ?O2 THEN Identify and test a vital COG possibility for a constrain The drive is ?O2 INFORMAL STRUCTURE FORMAL STRUCTURE

Slide 12

Learnable information portrayal Use of the question metaphysics as a deficient and advancing speculation pecking order. Conceivable rendition space (PVS) Use of conceivable adaptation spaces to speak to and utilize halfway learned information: Universe of Instances Plausible Upper Bound Concept Plausible Lower Bound Rules with PVS conditions Tasks with PVS conditions Object highlights with PVS idea Task highlights with PVS idea Feature Domain: PVS idea Range: PVS idea

Slide 13

Overview Research Problem, Approach, and Application Problem Solving Method: Task Reduction Learnable Knowledge Representation: Plausible Version Spaces Multistrategy Learning amid Problem Solving Agent Development Experiments Future Directions: Life-long Continuous Learning Teaching and Learning Demo Acknowledgments

Slide 14

Integrated displaying, learning and critical thinking Input Task Mixed-Initiative Problem Solving Ontology + Rules Generated Reduction Reject Reduction Accept Reduction Modeling Rule Specialization Specified Reduction Rule Generalization Rule Learning

Slide 15

2 Learns Rule_15 Which is an individual from Allied_Forces_1943 ? US_1943 Therefore we have to … Identify and test a vital COG contender for US_1943 We have to 3 5 Identify and test a key COG applicant comparing to an individual from the European_Axis_1943 Applies Rule_15 Which is an individual from European_Axis_1943 ? Refines ? Rule_15 4 Germany_1943 Therefore we have to Identify and test a vital COG possibility for Germany_1943 Accepts the illustration Disciple utilizes the scholarly standards in critical thinking, and refines them in view of master\'s input. We have to 1 Identify and test a vital COG hopeful comparing to an individual from the Allied_Forces_1943 Provides a case Learning Modeling Problem Solving Refining

Slide 16

Rule learning technique Analogy and Hint Guided Explanation Analogy-based Generalization Plausible adaptation space control conceivable clarifications PUB direction, clues Example of an errand decrease step PLB Incomplete clarification similarity Knowledge Base

Slide 17

Find a clarification of why the case is right We have to Identify and test a vital COG applicant relating to an individual from the Allied_Forces_1943 Which is an individual from Allied_Forces_1943 ? US_1943 Therefore we have to Identify and test a vital COG contender for US_1943 The clarification is the most ideal guess of the question and the appropriate response, in the protest cosmology. has_as_member Allied_Forces_1943 US_1943

Slide 18

Generate the PVS govern We have to Identify and test a key COG applicant relating to an individual from a constrain The compel is Allied_Forces_1943 has_as_member Allied_Forces_1943 US_1943 Therefore we have to Identify and test a vital COG contender for a drive The constrain is US_1943 IF Identify and test a key COG competitor comparing to an individual from a compel The drive is ?O1 Rewrite as clarification ?O1 has_as_member ?O2 Most broad speculation Plausible Upper Bound Condition ?O1 is multi_member_force has_as_member ?O2 is force Condition ?O1 is Allied_Forces_1943 has_as_member ?O2 ?O2 is US_1943 Plausible Lower Bound Condition ?O1 is equal_partners_multi_state_alliance has_as_member ?O2 ?O2 is single_state_force Most spec

Recommended
View more...