DOCSLIB.ORG

Predicate Logic: Semantics, Interpretations and Environments

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Predicate Logic: Semantics, Interpretations and Environments Warm-Up Problem Determine 훼[(푦 − 푧)/푥] where 훼 =def (∃푧 (푃 (푧) → (푄(푥, 푦) ∧ (∀푦 (푃 (푦) ∨ (∃푥 푃 (푥))))))) with 푃 (1) and 푄(2) predicates and 푥, 푦, 푧 variables. 1/48 Predicate Logic: Semantics, Interpretations and Environments Carmen Bruni Lecture 11 Based on slides by Jonathan Buss, Lila Kari, Anna Lubiw and Steve Wolfman with thanks to B. Bonakdarpour, A. Gao, D. Maftuleac, C. Roberts, R. Trefler, and P. Van Beek . Predicate Logic Semantics 2/48 Last Time • Did another substitution example (Please Review!) • Discussed Interpretations with respect to Predicate Logic . Predicate Logic Semantics 3/48 Learning Goals • Define an interpretation and an environment. • Give examples of interpretations and environments in specific situations. • Define validity, satisfiable and unsatisfiable. Predicate Logic Semantics 4/48 Leading Question Given a well-formed Predicate logic formula, is it T or F in some context? • In Propositional logic, a truth valuation was enough to assign a meaning to our atoms (propositional variables) • In Predicate logic, we need a lot more. Predicate Logic Semantics 5/48 However, if we consider an interpretation 풥 satisfying 푃 (푥, 푥) is 푥 is greater than 푥 and still consider the domain 풟 = {1}, then 훼 is false under this interpretation. We will formalize the notation of interpretations (and later environments) and explain what it means for a Predicate logic formula to be valid, satisfiable, and unsatisfiable. Motivating Example For example, if we consider the formula (for 푃 (2) a predicate symbol and variable 푥) 훼 =def (∀푥 푃 (푥, 푥)) and we use an interpretation ℐ of 푃 (푥, 푥) to mean 푥 is equal to 푥 and consider a domain 풟 = {1}, then indeed, 훼 is true under this interpretation. Predicate Logic Semantics 6/48 We will formalize the notation of interpretations (and later environments) and explain what it means for a Predicate logic formula to be valid, satisfiable, and unsatisfiable. Motivating Example For example, if we consider the formula (for 푃 (2) a predicate symbol and variable 푥) 훼 =def (∀푥 푃 (푥, 푥)) and we use an interpretation ℐ of 푃 (푥, 푥) to mean 푥 is equal to 푥 and consider a domain 풟 = {1}, then indeed, 훼 is true under this interpretation. However, if we consider an interpretation 풥 satisfying 푃 (푥, 푥) is 푥 is greater than 푥 and still consider the domain 풟 = {1}, then 훼 is false under this interpretation. Predicate Logic Semantics 6/48 Motivating Example For example, if we consider the formula (for 푃 (2) a predicate symbol and variable 푥) 훼 =def (∀푥 푃 (푥, 푥)) and we use an interpretation ℐ of 푃 (푥, 푥) to mean 푥 is equal to 푥 and consider a domain 풟 = {1}, then indeed, 훼 is true under this interpretation. However, if we consider an interpretation 풥 satisfying 푃 (푥, 푥) is 푥 is greater than 푥 and still consider the domain 풟 = {1}, then 훼 is false under this interpretation. We will formalize the notation of interpretations (and later environments) and explain what it means for a Predicate logic formula to be valid, satisfiable, and unsatisfiable. Predicate Logic Semantics 6/48 Brief Definition An interpretation consists of a domain as well as meanings for all of the constant, function and predicate symbols. Huth and Ryan use the term “model” instead of “interpretation”. More formally... Predicate Logic Semantics 7/48 Semantics: Interpretations Definition: Fix a set ℒ of constant symbols, function symbols, variable symbols and predicate symbols. (The “language” of our formulas.) An interpretation ℐ (for the set ℒ) consists of • A non-empty set dom(ℐ) or 풟ℐ or more simply 풟 , called the domain (or universe) of ℐ. • For each constant symbol 푐, a member 푐ℐ of dom(ℐ). • For each function symbol 푓(푖), an 푖-ary function 푓ℐ. • For each predicate symbol 푃 (푖), an 푖-ary predicate (relation) 푃 ℐ. When there are no variables and no quantifiers, this is more than enough to specify meaning to a formula. Predicate Logic Semantics 8/48 High Brow Comment Technically, our language should have all of the variable symbols we will ever need. In practice this is a bit cumbersome so we will usually forgo including variables in our language explicitly and simply use them as they appear in our formulas. Predicate Logic Semantics 9/48 Values of Variable-Free Terms Definition: Fix an interpretation ℐ. For each term 푡 containing no variables, the value of 푡 under interpretation ℐ, denoted 푡ℐ, is as follows. • If 푡 is a constant 푐, the value 푡ℐ is 푐ℐ. ℐ ℐ ℐ ℐ • If 푡 is 푓(푡1, … , 푡푛), the value 푡 is 푓 (푡1, … , 푡푛). The value of a term is always a member of the domain of ℐ. For example, consider 푓 a unary function and 0 a constant. If we have an interpretation ℐ with domain ℕ, 0ℐ the usual zero and 푓ℐ the usual successor function (increment by 1), then 푓(0)ℐ = 푓ℐ(0ℐ) = 1 . Predicate Logic Semantics 10/48 Watch out! Notice that in the previous example, even though we used the constant symbol 0, we still needed to specify that the interpretation of 0 is indeed the usual zero. For example, consider 푓 a unary function and 0 a constant. If we have an interpretation ℐ with domain ℕ, 0ℐ to be the usual number 1 and 푓ℐ the usual successor function (increment by 1), then 푓(0)ℐ = 푓ℐ(0ℐ) = 2 While doable, this is not advised... Predicate Logic Semantics 11/48 Clarity Most of these issues are taken care of by not using a symbol that could be misinterpreted as being in the domain. For example, consider 푓 a unary function and 푎 a constant. If we have an interpretation ℐ with domain ℕ, 푎ℐ to be the usual number 1 and 푓ℐ the usual successor function (increment by 1), then 푓(푎)ℐ = 푓ℐ(푎ℐ) = 2 . Predicate Logic Semantics 12/48 Total Function Another issue arises that your function’s interpretation must be defined on the entire domain! for a function with arity 푘, we need to define an interpretation such that the function 푓ℐ satisfies: 푓ℐ ∶ 풟푘 → 풟 that is, every 푘-tuple from the domain must map into the domain. Such functions capable of doing this are called total functions. • For example, the usual addition over the natural numbers is total since the sum of any two natural numbers gives another natural number. • However, the usual subtraction over the natural numbers is not total. For example, we cannot perform 2 − 6 and get a natural number. • Similarly, square roots over the integers (or even the real numbers!) is not a total function since the square root of −1 is not an integer (or a real number). Predicate Logic Semantics 13/48 Formulas with Variable-Free Terms Formulas get values in much the same fashion as terms, except that values of formulas lie in {T, F}. Definition: Fix an interpretation ℐ. For each formula 훼 containing no variables, the value of 훼 under interpretation ℐ, denoted 훼ℐ, is as follows. • If 훼 is 푃 (푡1, … , 푡푛), then {⎧T if ⟨푡ℐ , … , 푡ℐ ⟩ ∈ 푃 ℐ 훼ℐ = 1 푛 ⎨ ⎩{F otherwise. • If 훼 is (¬훽) or (훽 ⋆ 훾), then 훼ℐ is determined by 훽ℐ and 훾ℐ in the same way as for propositional logic. Predicate Logic Semantics 14/48 Example Let 푓(1) and ℎ(2) be function symbols. Let 푃 (1) and 푄(2) be predicate symbols and let 푎, 푏, 푐 be constant symbols. Define an interpretation by: • Domain: 풟 = {1, 2, 3} • Constants: 푎ℐ = 1, 푏ℐ = 2, 푐ℐ = 3 • Functions: 푓ℐ ∶ 푓ℐ(1) = 2, 푓ℐ(2) = 3, 푓ℐ(3) = 1 • ℎℐ ∶ (푥, 푦) ↦ min{푥, 푦} (min is the minimum function) • Predicates: 푃 ℐ = {1, 3} • 푄ℐ = {⟨1, 2⟩, ⟨3, 3⟩, ⟨3, 1⟩} What is the meaning of each of these formulas in this interpretation? • 푓(ℎ(푓(푎), 푓(푐)))ℐ • 푓(ℎ(푏, 푓(푎)))ℐ • 푄(푓(푐), 푎)ℐ ℐ • 푃 (ℎ(푓(푎), 푓(푐))) . Predicate Logic Semantics 15/48 Notice that 푓(푐)ℐ = 1 and 푎ℐ = 1. So if we set 풥 = ℐ except we define 푄풥 = {⟨1, 1⟩} then we see that ⟨푓풥 (푐풥 ), 푎풥 ⟩ ∈ 푄풥 . Follow Ups We saw that under ℐ we have that ⟨푓ℐ(푐ℐ), 푎ℐ⟩ ∉ 푄ℐ. Is there another interpretation 풥 such that ⟨푓풥 (푐풥 ), 푎풥 ⟩ ∈ 푄풥 ? . Predicate Logic Semantics 16/48 Follow Ups We saw that under ℐ we have that ⟨푓ℐ(푐ℐ), 푎ℐ⟩ ∉ 푄ℐ. Is there another interpretation 풥 such that ⟨푓풥 (푐풥 ), 푎풥 ⟩ ∈ 푄풥 ? Notice that 푓(푐)ℐ = 1 and 푎ℐ = 1. So if we set 풥 = ℐ except we define 푄풥 = {⟨1, 1⟩} then we see that ⟨푓풥 (푐풥 ), 푎풥 ⟩ ∈ 푄풥 . Predicate Logic Semantics 16/48 Notice that 푓(푐)ℐ = 1 and 푓(푎)ℐ = 2 and so ℎ(푓(푎), 푓(푐))ℐ = 1. Thus, if we set 풦 = ℐ except we define 푃 풦 = ∅ so that 1 ∉ 푃 풦, we see that 푃 (ℎ(푓(푎), 푓(푐)))풦 = F. Follow Ups We saw that under ℐ we have that 푃 (ℎ(푓(푎), 푓(푐)))ℐ = T. Is there another interpretation 풦 such that 푃 (ℎ(푓(푎), 푓(푐)))풦 = F? . Predicate Logic Semantics 17/48 Follow Ups We saw that under ℐ we have that 푃 (ℎ(푓(푎), 푓(푐)))ℐ = T. Is there another interpretation 풦 such that 푃 (ℎ(푓(푎), 푓(푐)))풦 = F? Notice that 푓(푐)ℐ = 1 and 푓(푎)ℐ = 2 and so ℎ(푓(푎), 푓(푐))ℐ = 1.
Load more

Recommended publications
  • “The Church-Turing “Thesis” As a Special Corollary of Gödel's
    “The Church-Turing “Thesis” as a Special Corollary of Gödel’s Completeness Theorem,” in Computability: Turing, Gödel, Church, and Beyond, B. J. Copeland, C. Posy, and O. Shagrir (eds.), MIT Press (Cambridge), 2013, pp. 77-104. Saul A. Kripke This is the published version of the book chapter indicated above, which can be obtained from the publisher at https://mitpress.mit.edu/books/computability. It is reproduced here by permission of the publisher who holds the copyright. © The MIT Press The Church-Turing “ Thesis ” as a Special Corollary of G ö del ’ s 4 Completeness Theorem 1 Saul A. Kripke Traditionally, many writers, following Kleene (1952) , thought of the Church-Turing thesis as unprovable by its nature but having various strong arguments in its favor, including Turing ’ s analysis of human computation. More recently, the beauty, power, and obvious fundamental importance of this analysis — what Turing (1936) calls “ argument I ” — has led some writers to give an almost exclusive emphasis on this argument as the unique justification for the Church-Turing thesis. In this chapter I advocate an alternative justification, essentially presupposed by Turing himself in what he calls “ argument II. ” The idea is that computation is a special form of math- ematical deduction. Assuming the steps of the deduction can be stated in a first- order language, the Church-Turing thesis follows as a special case of G ö del ’ s completeness theorem (first-order algorithm theorem). I propose this idea as an alternative foundation for the Church-Turing thesis, both for human and machine computation. Clearly the relevant assumptions are justified for computations pres- ently known.
    [Show full text]
  • A Compositional Analysis for Subset Comparatives∗ Helena APARICIO TERRASA–University of Chicago
    A Compositional Analysis for Subset Comparatives∗ Helena APARICIO TERRASA–University of Chicago Abstract. Subset comparatives (Grant 2013) are amount comparatives in which there exists a set membership relation between the target and the standard of comparison. This paper argues that subset comparatives should be treated as regular phrasal comparatives with an added presupposi- tional component. More specifically, subset comparatives presuppose that: a) the standard has the property denoted by the target; and b) the standard has the property denoted by the matrix predi- cate. In the account developed below, the presuppositions of subset comparatives result from the compositional principles independently required to interpret those phrasal comparatives in which the standard is syntactically contained inside the target. Presuppositions are usually taken to be li- censed by certain lexical items (presupposition triggers). However, subset comparatives show that presuppositions can also arise as a result of semantic composition. This finding suggests that the grammar possesses more than one way of licensing these inferences. Further research will have to determine how productive this latter strategy is in natural languages. Keywords: Subset comparatives, presuppositions, amount comparatives, degrees. 1. Introduction Amount comparatives are usually discussed with respect to their degree or amount interpretation. This reading is exemplified in the comparative in (1), where the elements being compared are the cardinalities corresponding to the sets of books read by John and Mary respectively: (1) John read more books than Mary. |{x : books(x) ∧ John read x}| ≻ |{y : books(y) ∧ Mary read y}| In this paper, I discuss subset comparatives (Grant (to appear); Grant (2013)), a much less studied type of amount comparative illustrated in the Spanish1 example in (2):2 (2) Juan ha leído más libros que El Quijote.
    [Show full text]
  • Interpretations and Mathematical Logic: a Tutorial
    Interpretations and Mathematical Logic: a Tutorial Ali Enayat Dept. of Philosophy, Linguistics, and Theory of Science University of Gothenburg Olof Wijksgatan 6 PO Box 200, SE 405 30 Gothenburg, Sweden e-mail: [email protected] 1. INTRODUCTION Interpretations are about `seeing something as something else', an elusive yet intuitive notion that has been around since time immemorial. It was sys- tematized and elaborated by astrologers, mystics, and theologians, and later extensively employed and adapted by artists, poets, and scientists. At first ap- proximation, an interpretation is a structure-preserving mapping from a realm X to another realm Y , a mapping that is meant to bring hidden features of X and/or Y into view. For example, in the context of Freud's psychoanalytical theory [7], X = the realm of dreams, and Y = the realm of the unconscious; in Khayaam's quatrains [11], X = the metaphysical realm of theologians, and Y = the physical realm of human experience; and in von Neumann's foundations for Quantum Mechanics [14], X = the realm of Quantum Mechanics, and Y = the realm of Hilbert Space Theory. Turning to the domain of mathematics, familiar geometrical examples of in- terpretations include: Khayyam's interpretation of algebraic problems in terms of geometric ones (which was the remarkably innovative element in his solution of cubic equations), Descartes' reduction of Geometry to Algebra (which moved in a direction opposite to that of Khayyam's aforementioned interpretation), the Beltrami-Poincar´einterpretation of hyperbolic geometry within euclidean geom- etry [13]. Well-known examples in mathematical analysis include: Dedekind's interpretation of the linear continuum in terms of \cuts" of the rational line, Hamilton's interpertation of complex numbers as points in the Euclidean plane, and Cauchy's interpretation of real-valued integrals via complex-valued ones using the so-called contour method of integration in complex analysis.
    [Show full text]
  • Church's Thesis and the Conceptual Analysis of Computability
    Church’s Thesis and the Conceptual Analysis of Computability Michael Rescorla Abstract: Church’s thesis asserts that a number-theoretic function is intuitively computable if and only if it is recursive. A related thesis asserts that Turing’s work yields a conceptual analysis of the intuitive notion of numerical computability. I endorse Church’s thesis, but I argue against the related thesis. I argue that purported conceptual analyses based upon Turing’s work involve a subtle but persistent circularity. Turing machines manipulate syntactic entities. To specify which number-theoretic function a Turing machine computes, we must correlate these syntactic entities with numbers. I argue that, in providing this correlation, we must demand that the correlation itself be computable. Otherwise, the Turing machine will compute uncomputable functions. But if we presuppose the intuitive notion of a computable relation between syntactic entities and numbers, then our analysis of computability is circular.1 §1. Turing machines and number-theoretic functions A Turing machine manipulates syntactic entities: strings consisting of strokes and blanks. I restrict attention to Turing machines that possess two key properties. First, the machine eventually halts when supplied with an input of finitely many adjacent strokes. Second, when the 1 I am greatly indebted to helpful feedback from two anonymous referees from this journal, as well as from: C. Anthony Anderson, Adam Elga, Kevin Falvey, Warren Goldfarb, Richard Heck, Peter Koellner, Oystein Linnebo, Charles Parsons, Gualtiero Piccinini, and Stewart Shapiro. I received extremely helpful comments when I presented earlier versions of this paper at the UCLA Philosophy of Mathematics Workshop, especially from Joseph Almog, D.
    [Show full text]
  • How to Build an Interpretation and Use Argument
    Timothy S. Brophy Professor and Director, Institutional Assessment How to Build an University of Florida Email: [email protected] Interpretation and Phone: 352‐273‐4476 Use Argument To review the concept of validity as it applies to higher education Today’s Goals Provide a framework for developing an interpretation and use argument for assessments Validity What it is, and how we examine it • Validity refers to the degree to which evidence and theory support the interpretations of the test scores for proposed uses of tests. (p. 11) defined Source: • American Educational Research Association (AERA), American Psychological Association (APA), & National Council on Measurement in Education (NCME). (2014). Standards for educational and psychological testing. Validity Washington, DC: AERA The Importance of Validity The process of validation Validity is, therefore, the most involves accumulating relevant fundamental consideration in evidence to provide a sound developing tests and scientific basis for the proposed assessments. score and assessment results interpretations. (p. 11) Source: AERA, APA, & NCME. (2014). Standards for educational and psychological testing. Washington, DC: AERA • Most often this is qualitative; colleagues are a good resource • Review the evidence How Do We • Common sources of validity evidence Examine • Test Content Validity in • Construct (the idea or theory that supports the Higher assessment) • The validity coefficient Education? Important distinction It is not the test or assessment itself that is validated, but the inferences one makes from the measure based on the context of its use. Therefore it is not appropriate to refer to the ‘validity of the test or assessment’ ‐ instead, we refer to the validity of the interpretation of the results for the test or assessment’s intended purpose.
    [Show full text]
  • Interpreting Arguments
    Informal Logic IX.1, Winter 1987 Interpreting Arguments JONATHAN BERG University of Haifa We often speak of "the argument in" Two notions which would merit ex­ a particular piece of discourse. Although tensive analysis in a broader context will such talk may not be essential to only get a bit of clarification here. By philosophy in the philosophical sense, 'claims' I mean what are variously call­ it is surely a practical requirement for ed 'propositions', 'statements', or 'con­ most work in philosophy, whether in the tents'. I remain neutral with regard to interpretation of classical texts or in the their ontological status and shall not pre­ discussion of contemporary work, and tend to know exactly how to individuate it arises as well in everyday contexts them, especially in relation to sentences. wherever we care about reasons given I shall assume that for every claim there in support of a claim. (I assume this is is at least one literal, declarative why developing the ability to identify sentence typically used for making that arguments in discourse is a major goal claim, and often I shall not distinguish in many courses in informal logic or between the claim, proper, and such a critical thinking-the major goal when I sentence. teach the subject.) Thus arises the ques­ What I mean when I say that one tion, just how do we get from text to claim follows from other claims is much argument? harder to specify than what I do not I shall propose a set of principles for mean. I certainly do not have in mind the extrication of argu ments from any mere psychological or causal con­ argumentative prose.
    [Show full text]
  • Why We Should Abandon the Semantic Subset Principle
    LANGUAGE LEARNING AND DEVELOPMENT https://doi.org/10.1080/15475441.2018.1499517 Why We Should Abandon the Semantic Subset Principle Julien Musolinoa, Kelsey Laity d’Agostinob, and Steve Piantadosic aDepartment of Psychology and Center for Cognitive Science, Rutgers University, Piscataway, NJ, USA; bDepartment of Philosophy, Rutgers University, New Brunswick, USA; cDepartment of Brain and Cognitive Sciences, University of Rochester, Rochester, USA ABSTRACT In a recent article published in this journal, Moscati and Crain (M&C) show- case the explanatory power of a learnability constraint called the Semantic Subset Principle (SSP) (Crain et al. 1994). If correct, M&C’s argument would represent a compelling demonstration of the operation of an innate, domain specific, learning principle. However, in trying to make the case for the SSP, M&C fail to clearly define their hypothesis, omit discussion of key facts, and contradict the theory itself. Moreover, their presentation does not address the core arguments and alternatives against their account available in the literature and misrepresents the position of their critics. Once these shortcomings are understood, a very different conclusion emerges: its historical significance notwithstanding, the SSP represents a flawed and obsolete account that should be abandoned. We discuss the implications of the demise of the SSP and describe more powerful and plausible alternatives that provide a better foundation for ongoing work in language acquisition. Introduction A common assumption within generative models of language acquisition is that learners are guided by a built-in constraint called the Subset Principle (SP) (Baker, 1979; Pinker, 1979; Dell, 1981; Berwick, 1985; Manzini & Wexler, 1987; among others).
    [Show full text]
  • A Description and Interpretation of the NPGS Birdsfoot Trefoil Core Subset Collection
    A Description and Interpretation of the NPGS Birdsfoot Trefoil Core Subset Collection J. J. Steiner,* P. R. Beuselinck, S. L. Greene, J. A Kamm, J. H. Kirkbride, and C. A. Roberts ABSTRACT active collection. This should increase evaluation effi- Systematic evaluations for a range of traits from accessions in ciency, reduce the number of accession requests, and re- National Plant Germplasm System (NPGS) core subset collections duce the frequency and costs for collection regeneration should assist collection management and enhance germplasm utiliza- (Greene and McFerson, 1994). Well characterized collec- tion. The objectives of this research were to: (i) evaluate the 48-accession tions can help plant explorers develop plans for collect- NPGS birdsfoot trefoil (Lotus corniculatus L.) core subset collection ing unique genetic materials that are not represented by by means of a variety of biochemical, morphological, and agronomic present ex situ collection holdings (Steiner and Greene, characters; (ii) determine how these characters were distributed among 1996; Greene et al., 1999a,b). the core subset accessions and associations among plant and ecogeo- In the past, most core subset collections were based graphic characters; (iii) define genetic diversity pools on the basis of on just a few traits or the geographic origins of the acces- descriptor interpretive groups; and (iv) develop a method to utilize the core subset as a reference collection to evaluate newly acquired acces- sions. Assessing a range of collection descriptors, includ- sions for their similarity or novelty. Geographic information system ing plant morphology, biochemistry, and collecting site (GIS) databases were used to estimate the ecogeography of accession habitat, has advantages that may be lost when only a sin- origins.
    [Show full text]
  • On the No-Counterexample Interpretation Basic Research in Computer Science
    BRICS Basic Research in Computer Science BRICS RS-97-42 U. Kohlenbach: On the No-Counterexample Interpretation On the No-Counterexample Interpretation Ulrich Kohlenbach BRICS Report Series RS-97-42 ISSN 0909-0878 December 1997 Copyright c 1997, BRICS, Department of Computer Science University of Aarhus. All rights reserved. Reproduction of all or part of this work is permitted for educational or research use on condition that this copyright notice is included in any copy. See back inner page for a list of recent BRICS Report Series publications. Copies may be obtained by contacting: BRICS Department of Computer Science University of Aarhus Ny Munkegade, building 540 DK–8000 Aarhus C Denmark Telephone: +45 8942 3360 Telefax: +45 8942 3255 Internet: [email protected] BRICS publications are in general accessible through the World Wide Web and anonymous FTP through these URLs: http://www.brics.dk ftp://ftp.brics.dk This document in subdirectory RS/97/42/ On the no-counterexample interpretation Ulrich Kohlenbach BRICS∗ Department of Computer Science University of Aarhus Ny Munkegade, Bldg. 540 DK-8000 Aarhus C, Denmark [email protected] December 1997 Abstract In [15],[16] Kreisel introduced the no-counterexample interpretation (n.c.i.) of Peano arithmetic. In particular he proved, using a complicated ε-substitution method (due to W. Ackermann), that for every theorem A (A prenex) of first-order Peano arithmetic PA one can find ordinal recursive functionals ΦA of order type <ε0 which realize the Herbrand normal form AH of A. Subsequently more perspicuous proofs of this fact via functional interpretation (com- bined with normalization) and cut-elimination where found.
    [Show full text]
  • Lambda Calculus and Computation 6.037 – Structure and Interpretation of Computer Programs
    Lambda Calculus and Computation 6.037 { Structure and Interpretation of Computer Programs Benjamin Barenblat [email protected] Massachusetts Institute of Technology With material from Mike Phillips, Nelson Elhage, and Chelsea Voss January 30, 2019 Benjamin Barenblat 6.037 Lambda Calculus and Computation : Build a calculating machine that gives a yes/no answer to all mathematical questions. Figure: Alonzo Church Figure: Alan Turing (1912-1954), (1903-1995), lambda calculus Turing machines Theorem (Church, Turing, 1936): These models of computation can't solve every problem. Proof: next! Limits to Computation David Hilbert's Entscheidungsproblem (1928) Benjamin Barenblat 6.037 Lambda Calculus and Computation Figure: Alonzo Church Figure: Alan Turing (1912-1954), (1903-1995), lambda calculus Turing machines Theorem (Church, Turing, 1936): These models of computation can't solve every problem. Proof: next! Limits to Computation David Hilbert's Entscheidungsproblem (1928): Build a calculating machine that gives a yes/no answer to all mathematical questions. Benjamin Barenblat 6.037 Lambda Calculus and Computation Theorem (Church, Turing, 1936): These models of computation can't solve every problem. Proof: next! Limits to Computation David Hilbert's Entscheidungsproblem (1928): Build a calculating machine that gives a yes/no answer to all mathematical questions. Figure: Alonzo Church Figure: Alan Turing (1912-1954), (1903-1995), lambda calculus Turing machines Benjamin Barenblat 6.037 Lambda Calculus and Computation Proof: next! Limits to Computation David Hilbert's Entscheidungsproblem (1928): Build a calculating machine that gives a yes/no answer to all mathematical questions. Figure: Alonzo Church Figure: Alan Turing (1912-1954), (1903-1995), lambda calculus Turing machines Theorem (Church, Turing, 1936): These models of computation can't solve every problem.
    [Show full text]
  • Quantum Set Theory Extending the Standard Probabilistic Interpretation of Quantum Theory (Extended Abstract)
    Quantum Set Theory Extending the Standard Probabilistic Interpretation of Quantum Theory (Extended Abstract) Masanao Ozawa∗ Graduate School of Information Science, Nagoya University Nagoya, Japan [email protected] The notion of equality between two observables will play many important roles in foundations of quantum theory. However, the standard probabilistic interpretation based on the conventional Born formula does not give the probability of equality relation for a pair of arbitrary observables, since the Born formula gives the probability distribution only for a commuting family of observables. In this paper, quantum set theory developed by Takeuti and the present author is used to systematically extend the probabilistic interpretation of quantum theory to define the probability of equality relation for a pair of arbitrary observables. Applications of this new interpretation to measurement theory are discussed briefly. 1 Introduction Set theory provides foundations of mathematics. All the mathematical notions like numbers, functions, relations, and structures are defined in the axiomatic set theory, ZFC (Zermelo-Fraenkel set theory with the axiom of choice), and all the mathematical theorems are required to be provable in ZFC [15]. Quan- tum set theory instituted by Takeuti [14] and developed by the present author [11] naturally extends the logical basis of set theory from classical logic to quantum logic [1]. Accordingly, quantum set theory extends quantum logical approach to quantum foundations from propositional logic to predicate logic and set theory. Hence, we can expect that quantum set theory will provide much more systematic inter- pretation of quantum theory than the conventional quantum logic approach [3]. The notion of equality between quantum observables will play many important roles in foundations of quantum theory, in particular, in the theory of measurement and disturbance [9, 10].
    [Show full text]
  • An Information-Theoretic Perspective on Model Interpretation
    Learning to Explain: An Information-Theoretic Perspective on Model Interpretation Jianbo Chen 1 2 Le Song 3 4 Martin J. Wainwright 1 Michael I. Jordan 1 Abstract chine learning is feature selection, which selects a subset of We introduce instancewise feature selection as features that are useful to build a good predictor for a spec- a methodology for model interpretation. Our ified response variable (Guyon & Elisseeff, 2003). While method is based on learning a function to ex- feature selection produces a global importance of features tract a subset of features that are most informative with respect to the entire labeled data set, instancewise fea- for each given example. This feature selector is ture selection measures feature importance locally for each trained to maximize the mutual information be- instance labeled by the model. tween selected features and the response variable, Existing work on interpreting models approach the problem where the conditional distribution of the response from two directions. The first line of work computes the variable given the input is the model to be ex- gradient of the output of the correct class with respect to the plained. We develop an efficient variational ap- input vector for the given model, and uses it as a saliency proximation to the mutual information, and show map for masking the input (Simonyan et al., 2013; Springen- the effectiveness of our method on a variety of berg et al., 2014). The gradient is computed using a Parzen synthetic and real data sets using both quantitative window approximation of the original classifier if the origi- metrics and human evaluation.
    [Show full text]