DOCSLIB.ORG

Belief Revision in Dynamic Abducers Through Meta-Abduction THESIS

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Belief Revision in Dynamic Abducers Through Meta-Abduction THESIS Belief Revision in Dynamic Abducers through Meta-Abduction THESIS Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University By Vivek Bharathan Graduate Program in Computer Science and Engineering The Ohio State University 2010 Master's Examination Committee: Dr. John Josephson, Advisor Professor B. Chandrasekaran, Advisor Copyright by Vivek Bharathan 2010 Abstract Abduction machines (or abducers) infer to the best explanation for the data presented to them, and may accumulate beliefs (the conclusions of the inferences) about the world. An abducer‟s beliefs are justified as being the best explanation in contrast with alternative hypotheses. However, the best explanation available to (achievable by) an abducer need not be the true explanation for a variety of reasons including: incomplete search for alternative explanations, insufficient data, and inadequate background knowledge for evaluating alternatives. In light of this fallibility, algorithms were investigated for detecting and correcting errors, for dynamic abducers, by comparing a range of alternative algorithms with regard to effectiveness and computational costs. Dynamic abducers interpret an incoming information stream by producing their best explanations at any point in time. They accumulate new beliefs, and update older beliefs, in the light of new information. In the present work, algorithms were developed for detecting and correcting errors in the accumulated beliefs of such abducers. These algorithms treat the problem of identifying errors as meta-abduction, where certain anomalies that occur during processing are explained as resulting from specific mistakes in previous abductive processing. Errors are then corrected, and beliefs revised, by adopting alternative explanations. A brute-force algorithm for this meta-abduction is computationally intractable, so heuristics were developed with prospects of improving performance. Since a priori mathematical analysis of the algorithms using these heuristics did not reveal ii useful bounds, simulation experiments were conducted, using a specimen domain. The domain was that of multi-object tracking, where an abduction machine, over time, attempts to maintain the track history of mobile entities, based on sensor reports. The experimental results suggest that, using only the heuristics that were investigated, belief revision by meta-abduction enables only small improvements in correctness, and is computationally expensive. iii Vita 1997................................................................Padma Seshadri Bala Bhavan Senior Secondary School 2001................................................................B.E. Computer Science and Engineering, Sri Venkateswara College of Engineering 2002-2003 .....................................................Graduate Teaching Associate, Department of Computer Science and Engineering, The Ohio State University 2003-2008 .....................................................Graduate Research Associate, Department of Computer Science and Engineering, The Ohio State University Publications Vivek Bharathan and John. R. Josephson, "Belief Revision controlled by Meta- Abduction", Special Issue of Logic Journal of IGPL (Oxford University Press), Abduction, Practical Reasoning and Creative Inferences in Science, edited by L. Magnani, 2006 (Volume 14) Vivek Bharathan and John Josephson, “Detecting and Correcting Mistakes in Information iv Fusion”, in proceedings of The Army Science Conference, 2006 Vivek Bharathan and John Josephson, "An Abductive Framework for Level One Information Fusion", in proceedings of The 9th International Conference on Information Fusion, 2006 John Josephson and Vivek Bharathan, "Abductive Inference Engine for Entity Re- identification and Tracking", in proceedings of The Army Science Conference, 2004 Fields of Study Major Field: Computer Science and Engineering Specialization: Artificial Intelligence v Table of Contents Abstract ............................................................................................................................... ii Vita ..................................................................................................................................... iv Table of Contents ............................................................................................................... vi List of Tables ..................................................................................................................... ix List of Figures ..................................................................................................................... x Chapter 1: Introduction ...................................................................................................... 1 1.1 Dynamic Abduction .................................................................................................. 4 Chapter 2: Belief Revision in Dynamic Abducers ............................................................. 9 2.1 Detecting Errors in Dynamic Abducers .................................................................... 9 2.1.2 Trigger 2: Predictions Contradicted by Observations: ..................................... 10 2.1.3 Trigger 3: Clashing Explanations ..................................................................... 11 2.2 Correcting Errors in Dynamic Abducers ................................................................. 11 2.3 Problem Space Complexity for Belief Revision in Abductive Agents ................... 14 Chapter 3: The Abductive Approach to Multi-Object Tracking ....................................... 19 3.1 Description of the ASAS Domain ........................................................................... 19 vi 3.2 The Entity Re-identification Algorithm in Smart-ASAS ........................................ 21 3.3 An Illustrative Example of Abductive Tracking and Error Correction in Smart- ASAS ............................................................................................................................. 24 3.3.1 Initial State ........................................................................................................ 25 3.3.2 Processing Sensor Reports - Generation, Evaluation and Acceptance of Hypotheses................................................................................................................. 26 3.3.3 Revision of previous conclusions in Smart-ASAS ........................................... 28 Chapter 4: The Smart-ASAS Testbed ............................................................................... 33 4.1 The Class of Meta-Abductive Error-Correction Algorithms .................................. 36 4.1.1 The “In Denial” Algorithm ............................................................................... 37 4.1.2 The “Acknowledge Discrepancy” Algorithm ................................................... 38 4.1.3 The “Revision with Recency Heuristic” Algorithm ......................................... 39 4.1.4 The “Revision with Entrenchment Heuristic” Algorithm ................................ 40 4.2 Performance Measures for the Error-Correction Algorithms ................................. 42 Chapter 5: Experimental Results from the Simulated ASAS Domain ............................. 44 5.1 Performance of the Error-Correction Algorithms ................................................... 44 Chapter 6: Summary and Discussion ................................................................................ 69 References ......................................................................................................................... 71 vii Appendix A: An Illustrative Example of Abductive Tracking and Error Correction in Smart-ASAS with Alternative Revisions.......................................................................... 75 Appendix B: The PEIRCE Hypothesis Assembly Algorithm .......................................... 82 viii List of Tables Table 1. Meta-Abductive Belief Revision versus Abductive Processing ......................... 13 Table 2. Parameters in Complexity Analysis of Abductive Belief Revision .................... 16 ix List of Figures Figure 1. Intial State of the World .................................................................................... 25 Figure 2. Decision D and Second Report.......................................................................... 28 Figure 3. Decision D' and Third Report ............................................................................ 30 Figure 4. Final State Estimate ........................................................................................... 32 Figure 5. Performance Measure - Average Track Length ................................................ 46 Figure 6. Performance Measure - Number of Miscorrelations ......................................... 50 Figure 7. Performance Measure – Error Rate ................................................................... 52 Figure 8. Performance Measure – Efficiency .................................................................. 57 Figure 9. Actual Errors versus Observed Errors in different Revision Algorithms .......... 60 Figure 10. Performance of the BR with Recency Algorithm ............................................ 63 Figure 11. Performance of the BR with Entrenchment Algorithm ................................... 65 Figure 12. Initial State of the World ................................................................................
Load more

Recommended publications
  • Redalyc.Evaluation of Consistency of the Knowledge in Agents
    Acta Universitaria ISSN: 0188-6266 [email protected] Universidad de Guanajuato México Contreras, Meliza; Rodriguez, Miguel; Bello, Pedro; Aguirre, Rodolfo Evaluation of Consistency of the Knowledge in Agents Acta Universitaria, vol. 22, marzo, 2012, pp. 42-47 Universidad de Guanajuato Guanajuato, México Available in: http://www.redalyc.org/articulo.oa?id=41623190006 How to cite Complete issue Scientific Information System More information about this article Network of Scientific Journals from Latin America, the Caribbean, Spain and Portugal Journal's homepage in redalyc.org Non-profit academic project, developed under the open access initiative Universidad de Guanajuato * ∗ ∗ ∗ In recent years, many formalisms have been proposed in the Artificial Intelligence literature to model common- sense reasoning. So, the revision and transformation of knowledge is widely recognized as a key problem in knowledge representation and reasoning. Reasons for the importance of this topic are the facts that intelligent systems are gradually developed and refined, and that often the environment of an intelligent system is not static but changes over time [4, 9]. Belief revision studies reasoning with changing information. Traditionally, belief revision techniques have been expressed using classical logic. Recently, the study of knowledge revision has increased since it can be applied to several areas of knowledge. Belief revision is a system that contains a corpus of beliefs which can be changed to accommodate new knowledge that may be inconsistent with any previous beliefs. Assuming the new belief is correct, as many of the previous ones should be removed as necessary so that the new belief can be in- corporated into a consistent corpus.
    [Show full text]
  • Explanations, Belief Revision and Defeasible Reasoning
    Artificial Intelligence 141 (2002) 1–28 www.elsevier.com/locate/artint Explanations, belief revision and defeasible reasoning Marcelo A. Falappa a, Gabriele Kern-Isberner b, Guillermo R. Simari a,∗ a Artificial Intelligence Research and Development Laboratory, Department of Computer Science and Engineering, Universidad Nacional del Sur Av. Alem 1253, (B8000CPB) Bahía Blanca, Argentina b Department of Computer Science, LG Praktische Informatik VIII, FernUniversitaet Hagen, D-58084 Hagen, Germany Received 15 August 1999; received in revised form 18 March 2002 Abstract We present different constructions for nonprioritized belief revision, that is, belief changes in which the input sentences are not always accepted. First, we present the concept of explanation in a deductive way. Second, we define multiple revision operators with respect to sets of sentences (representing explanations), giving representation theorems. Finally, we relate the formulated operators with argumentative systems and default reasoning frameworks. 2002 Elsevier Science B.V. All rights reserved. Keywords: Belief revision; Change theory; Knowledge representation; Explanations; Defeasible reasoning 1. Introduction Belief Revision systems are logical frameworks for modelling the dynamics of knowledge. That is, how we modify our beliefs when we receive new information. The main problem arises when that information is inconsistent with the beliefs that represent our epistemic state. For instance, suppose we believe that a Ferrari coupe is the fastest car and then we found out that some Porsche cars are faster than any Ferrari cars. Surely, we * Corresponding author. E-mail addresses: [email protected] (M.A. Falappa), [email protected] (G. Kern-Isberner), [email protected] (G.R.
    [Show full text]
  • Redalyc.Abductive Reasoning: Challenges Ahead
    THEORIA. Revista de Teoría, Historia y Fundamentos de la Ciencia ISSN: 0495-4548 [email protected] Universidad del País Vasco/Euskal Herriko Unibertsitatea España ALISEDA, Atocha Abductive Reasoning: Challenges Ahead THEORIA. Revista de Teoría, Historia y Fundamentos de la Ciencia, vol. 22, núm. 3, 2007, pp. 261- 270 Universidad del País Vasco/Euskal Herriko Unibertsitatea Donostia-San Sebastián, España Available in: http://www.redalyc.org/articulo.oa?id=339730804001 How to cite Complete issue Scientific Information System More information about this article Network of Scientific Journals from Latin America, the Caribbean, Spain and Portugal Journal's homepage in redalyc.org Non-profit academic project, developed under the open access initiative Abductive Reasoning: Challenges Ahead Atocha ALISEDA BIBLID [0495-4548 (2007) 22: 60; pp. 261-270] ABSTRACT: The motivation behind the collection of papers presented in this THEORIA forum on Abductive reasoning is my book Abductive Reasoning: Logical Investigations into the Processes of Discovery and Explanation. These contributions raise fundamental questions. One of them concerns the conjectural character of abduction. The choice of a logical framework for abduction is also discussed in detail, both its inferential aspect and search strategies. Abduction is also analyzed as inference to the best explanation, as well as a process of epistemic change, both of which chal- lenge the argument-like format of abduction. Finally, the psychological question of whether humans reason abduc- tively according to the models proposed is also addressed. I offer a brief summary of my book and then comment on and respond to several challenges that were posed to my work by the contributors to this issue.
    [Show full text]
  • Two Approaches to Belief Revision*
    an agent learns the proposition E, her posterior belief set is B0 = B ?E.2 While AGM’s belief revision operator (∗) is defined in wholly qualitative terms, the Lockean belief revision operator (÷) is defined by way of a diachronic version of Foley’s [13] Lockean thesis. That is, Lockean revision requires that Two Approaches to Belief Revision* when an agent learns E, she adopts the posterior belief set, B0 = B÷E, such that B0(X) just in case her posterior credence in X is no less than the Lockean threshold, t. So, for an agent to perform Lockean revision, she will require some procedure for updating her prior credence function, b, to a posterior credence function, b0. As it will facilitate the simplest and most straightforward presentation, we will assume that credences are updated via conditionalization. So, when an agent with the prior credence function b learns E, she adopts the posterior b0(·) = b(· j E).3 However, this choice is Abstract made largely for convenience, since most (if not all) of our results will hold for any update of the prior b by E to the posterior b0 that satisfies the following two constraints: (i) b0(E) > b(E) and (ii) In this paper, we compare and contrast two methods for the revision of qualitative (viz., “full”) if b0(X) ≥ t, then b(E ⊃ X) ≥ t.4 beliefs. The first (“Bayesian”) method is generated by a simplistic diachronic Lockean thesis In the next section, we will introduce Lockean revision, and provide some initial results designed requiring coherence with the agent’s posterior credences after conditionalization.
    [Show full text]
  • Two Dogmas of Belief Revision
    Two Dogmas of Belief Revision Hans Rott Institute for Philosophy, University of Regensburg, 93040 Regensburg, Germany [email protected] Abstract The pap er attacks the widely held view that b elief revison theories, as they have b een studied in the past two decades, are founded on the Principle of Informational Economy. The principle comes in twoversions. According to the rst, an agent should, when accepting a new piece of information, aim at a minimal change of his previously held b eliefs. If there are di erentways to e ect the b elief change, then the agent should, according to he second version, give up those b eliefs that are least entrenched. Although widely pro claimed by b elief revision theorists, I argue that b oth versions of the principle are dogmas that are not (and p erhaps should not b e) adhered to. I substantiate this claim with two simple mathematical obser- vations, and I defend it against four p ossible ob jections that involve contractions, reconstructions, disp ositions, and truths. 1. Intro duction Quine's pap er \Two Dogmas of Empiricism" (1951) was ep o ch-making in philosophy. It sealed the fate of logical p ositivism by undermining two presumptions of classical as well as mo dern empiricism: The idea that there is a sharp distinction b etween analytical judgements (knowledge of meanings) and synthetical judgements (knowledge of facts), and the idea that every meaningful sentence can b e reduced to a construction up on observation rep orts. The interesting p oint for us is that Quine closes his pap er with a b eautiful section on \empiricism without the dogmas" which deals almost exclusively with a topic that would nowadays b e called theory change ,orbelief revision .Hepaints a picture of howwe should, and for the most part do, accommo date our scienti c heritage if we meet with recalcitrant exp eriences.
    [Show full text]
  • A Defense of Explanatory Coherentism December 13, 2013
    Reason & Explanation: A defense of explanatory coherentism Ted L. Poston University of South Alabama December 13, 2013 ii Contents Preface vii 1 Introduction 1 1.1 A brief history of coherentism . 2 1.2 Two traditional objections to coherentism . 7 1.2.1 The Input Objection . 7 1.2.2 Alternative Systems Objection . 9 1.3 Overview . 11 2 Epistemic Conservatism 17 2.1 The anti-conservative probability argument . 19 2.2 Conservative Justification & Warranted Assertion . 26 2.3 Conservatism & Autobiographical Epistemology . 28 2.4 The `Extra Boost' and Conversion objections . 32 2.4.1 The \Extra Boost" Objection . 32 2.4.2 Conversion objections . 33 2.5 Conservatism & the perspectival character of justification . 35 2.5.1 The argument from perspective . 36 2.5.2 Two challenges . 38 2.6 Conclusion . 40 3 Reasons without first philosophy 41 3.1 The Basic Reasons Dilemma . 42 3.1.1 The First Horn . 42 3.1.2 The Second Horn . 44 3.2 The argument against first philosophy . 45 3.2.1 The nature of basic reasons . 45 3.2.2 No First Philosophy . 48 3.3 Framework Reasons . 51 iii iv CONTENTS 3.4 Weak foundationalism & framework reasons . 56 3.5 Bergmann on foundationalism and epistemic circularity . 58 3.6 Conclusion . 62 4 Explanation & Justification 63 4.1 Explanation & its virtues . 64 4.1.1 Three arguments for primitiveness . 67 4.1.2 The virtues of explanation . 73 4.2 An explanationist theory of justification . 78 4.2.1 The goal . 78 4.2.2 The Ex-J account . 79 4.2.3 Ex-J & Mentalism .
    [Show full text]
  • Giving a Step-By-Step Reduction from SAT to TSP and Giving Some Remarks on Neil Tennant's Changes of Mind
    Faculty of Mathematics and Natural Sciences Giving a step-by-step reduction from SAT to TSP and giving some remarks on Neil Tennant's Changes of Mind Bachelor Thesis Mathematics July 2014 Student: M.M. Bronts Supervisors: Prof. dr. J. Top and Prof. dr. B.P. Kooi Abstract Karp wrote an article about 21 decision problems. He gave instances for poly- nomial reductions between these problems, but he did not prove that these instances actually worked. In this thesis we will first prove that some of these instances are indeed correct. Second we will give some remarks on a belief contraction problem given by Neil Tennant in his book Changes of Mind. He describes what happens in our mind when we change our mind and argues that this is an NP-complete decision problem. 2 Contents 1 Introduction4 2 Historical background5 3 Introductory definitions and decision problems7 3.1 The Turing machine............................7 3.2 Some important definitions........................7 3.3 Decision problems.............................8 4 The Boolean satisfiability problem 10 4.1 Propositional logic............................. 10 4.2 Polynomial functions over F2 ....................... 11 4.3 SAT..................................... 13 5 Complexity theory and reduction 15 5.1 Complexity theory............................. 15 5.2 Polynomial time reduction......................... 17 5.3 Example: 3-SAT is NP-complete..................... 19 6 Reduction of SAT to TSP 23 6.1 SAT CLIQUE.............................. 23 6.2 CLIQUE VC............................... 24 6.3 VC HC.................................. 25 6.4 HC TSP................................. 29 7 Remarks on Changes of Mind by Neil Tennant 31 7.1 Tennant's belief system.......................... 31 7.2 The contraction problem.........................
    [Show full text]
  • 'Belief-Revision, the Ramsey Test, Monotonicity, and the So-Called
    THE REVIEW OF SYMBOLIC LOGIC Volume 1, Number 4, December 2008 BELIEF-REVISION, THE RAMSEY TEST, MONOTONICITY, AND THE SO-CALLED IMPOSSIBILITY RESULTS NEIL TENNANT Department of Philosophy, The Ohio State University Abstract. Peter Gardenfors¨ proved a theorem purporting to show that it is impossible to adjoin to the AGM-postulates for belief-revision a principle of monotonicity for revisions. The principle of monotonicity in question is implied by the Ramsey test for conditionals. So Gardenfors’¨ result has been interpreted as demonstrating that it is impossible to combine the Ramsey test for conditionals with the basic postulates for rational belief-revision. It is shown here that this interpretation of Gardenfors’¨ result is unwarranted. A new diagnosis is offered of a methodological error made in the statement of the key principle of monotonicity. Crucial applications of this principle in Gardenfors’¨ proof require one to regard as revisions what are really expansions. If monotonicity is stated only for genuine revisions, then Gardenfors’¨ proof does not go through. Nor can it; for, when the monotonicity principle for revisions is correctly formulated, one can actually establish a contrary consistency result. This requires only a slight adjustment to the postulates of AGM-theory, in order to ensure that the three operations of expansion, contraction, and revision trichotomize the domain of theory- changes. It is further shown that being careful in this way about the proper domains of definition of the three operations of expansion, contraction, and revision also disposes of another, more direct, impossibility result, due to Arlo-Costa,´ that targets the Ramsey test. §1.
    [Show full text]
  • Incorporating Relevance in Epistemic States in Belief Revision
    Incorporating Relevance in Epistemic States in Belief Revision James Delgrande,1 Pavlos Peppas2,3 1 School of Computing Science, Simon Fraser University, Burnaby, BC V5A1S6, Canada 2 Centre for AI, School of Software, University of Technology Sydney, NSW 2007, Australia 3 Dept of Business Administration, University of Patras, Patras 26500, Greece [email protected], [email protected] Abstract mula for revision. Indeed, as information repositories grow progressively larger, it is crucial that a belief change be re- We present an account of relevance in belief revision where, stricted to a “local” portion of a knowledge base. Conse- intuitively, one wants to only consider the relevant part of an agent’s epistemic state in a revision. We assume that rele- quently, an account of relevance in belief change is crucial, vance is a domain-specific notion, and that (ir)relevance as- both conceptually as well as computationally. sertions are given as part of the agent’s epistemic state. Such In this paper we present an account of relevance in belief assertions apply in a given context, and are of the form “in the revision. We take relevance to be a domain dependent no- case that formula σ holds, the Y part of the agent’s epistemic tion, in which assertions of (ir)relevance are part of the back- state is independent of the rest of the epistemic state”, where ground theory of a domain. For example, we may have the Y is part of the signature of the language. Two approaches assertion that, in the context of a party, whether Alice is there are given, one in which (in semantic terms) conditions are is independent of whether Bob is there (because they don’t placed on a faithful ranking on possible worlds to enforce the know each other).
    [Show full text]
  • 1993-Abduction As Belief Revision: a Model of Preferred Explanations
    From: AAAI-93 Proceedings. Copyright © 1993, AAAI (www.aaai.org). All rights reserved. A s Craig Boutilier and Verhica Becher Department of Computer Science University of British Columbia Vancouver, British Columbia CANADA, V6T 172 email: cebly,[email protected] Abstract “logic-based” frameworks also. Peirce (see Rescher (1978)) We propose a natural model of abduction based on the re- discusses the “plausibility” of explanations, as do Quine and vision of the epistemic state of an agent. We require that Ullian (1970). Consistency-based diagnosis (Reiter 1987; explanations be sufficient to induce belief in an observation de Kleer, Mackworth and Reiter 1990) uses abnormality in a manner that adequately accounts for factual and hypo- assumptions to capture the context dependence of explana- thetical observations. Our model will generate explanations tions; and preferred explanations are those that minimize that nonmonotonically predict an observation, thus general- abnormalities. Poole’s (1989) assumption-based framework izing most current accounts, which require some deductive captures some of these ideas by explicitly introducing a set relationship between explanation and observation. It also of default assumptions to account for the nonmonotonicity provides a natural preference ordering on explanations, de- of explanations. fined in terms of normality or plausibility. We reconstruct We propose a semantic framework for abduction that cap- the Theorist system in our framework, and show how it can tures the spirit of probabilistic proposals, but in a qualitative be extended to accommodate our predictive explanations and fashion, and in such a way that existing logic-based proposals semantic preferences on explanations. can be represented as well.
    [Show full text]
  • Belief Revision Meets Philosophy of Science LOGIC, EPISTEMOLOGY, and the UNITY of SCIENCE
    Belief Revision Meets Philosophy of Science LOGIC, EPISTEMOLOGY, AND THE UNITY OF SCIENCE VOLUME 21 Editors Shahid Rahman, University of Lille III, France John Symons, University of Texas at El Paso, U.S.A. Editorial Board Jean Paul van Bendegem, Free University of Brussels, Belgium Johan van Benthem, University of Amsterdam, the Netherlands Jacques Dubucs, University of Paris I-Sorbonne, France Anne Fagot-Largeault Collège de France, France Bas van Fraassen, Princeton University, U.S.A. Dov Gabbay, King’s College London, U.K. Jaakko Hintikka, Boston University, U.S.A. Karel Lambert, University of California, Irvine, U.S.A. Graham Priest, University of Melbourne, Australia Gabriel Sandu, University of Helsinki, Finland Heinrich Wansing, Technical University Dresden, Germany Timothy Williamson, Oxford University, U.K. Logic, Epistemology, and the Unity of Science aims to reconsider the question of the unity of science in light of recent developments in logic. At present, no single logical, semantical or methodological framework dominates the philosophy of science. However, the editors of this series believe that formal techniques like, for example, independence friendly logic, dialogical logics, multimodal logics, game theoretic semantics and linear logics, have the potential to cast new light on basic issues in the discussion of the unity of science. This series provides a venue where philosophers and logicians can apply specific technical insights to fundamental philosophical problems. While the series is open to a wide variety of perspectives, including the study and analysis of argumentation and the critical discussion of the relationship between logic and the philosophy of science, the aim is to provide an integrated picture of the scientific enterprise in all its diversity.
    [Show full text]
  • The Learning Power of Belief Revision
    The Learning Power of Belief Revision Kevin T. Kelly Department of Philosophy Carnegie Mellon University [email protected] Abstract Belief revision theory aims to describe how one should change one’s beliefs when they are contradicted by newly input information. The guiding principle of belief revision theory is to change one’s prior beliefs as little as possible in order to maintain consistency with the new information. Learning theory focuses, instead, on learning power: the ability to arrive at true beliefs in a wide range of possible environments. The goal of this paper is to bridge the two approaches by providing a learning theoretic analysis of the learning power of belief revision methods proposed by Spohn, Boutilier, Darwiche and Pearl, and others. The results indicate that learning power depends sharply on details of the methods. Hence, learning power can provide a well-motivated constraint on the design and implementation of concrete belief revision methods. 1 Introduction Intelligent systems act on the basis of fallible, general beliefs, such as “My car always stays where I put it” or “rough objects of a given shape and size have more air resistance than smooth ones”. When general beliefs of this sort are refuted by new information, they must somehow be revised to accommodate it. Accordingly, there is increasing interdisciplinary interest in Belief revision theory, which aims to specify how to rationally revise refuted beliefs [1] [4] [10] [14] [32] [33] [5] [6] [36]. The official motivation for belief revision theory [7] is to minimize change or loss to one’s prior belief state when new, conflicting information is incorporated.
    [Show full text]