Formal Theories for Reasoning about Parthood, Connection, and Location

Slides:



Advertisements
Similar presentations
Automated Theorem Proving Lecture 1. Program verification is undecidable! Given program P and specification S, does P satisfy S?
Advertisements

1 First order theories (Chapter 1, Sections 1.4 – 1.5)
The Engineering Design of Systems: Models and Methods
CSE115/ENGR160 Discrete Mathematics 04/03/12 Ming-Hsuan Yang UC Merced 1.
From Layered Mereotopology to Dynamic Spatial Ontology Maureen Donnelly and Barry Smith Department of Philosophy, University at Buffalo and Institute for.
From Layered Mereotopology to Dynamic Spatial Ontology Maureen Donnelly and Barry Smith Department of Philosophy, University at Buffalo and Institute for.
CSE115/ENGR160 Discrete Mathematics 03/31/11
4. The Postulates of Quantum Mechanics 4A. Revisiting Representations
A Formal Theory for Spatial Representation and Reasoning in Biomedical Ontologies Maureen Donnelly Thomas Bittner.
Introduction to Logic for Artificial Intelligence Lecture 2 Erik Sandewall 2010.
Algebra Algebra – defined by the tuple:  A, o 1, …, o k ; R 1, …, R m ; c 1, …, c k  Where: A is a non-empty set o i is the function, o i : A p i  A.
Probability and Statistics Review Thursday Sep 11.
1 First order theories. 2 Satisfiability The classic SAT problem: given a propositional formula , is  satisfiable ? Example:  Let x 1,x 2 be propositional.
Chapter 4: A Universal Program 1. Coding programs Example : For our programs P we have variables that are arranged in a certain order: Y 1 X 1 Z 1 X 2.
Relations Chapter 9.
Induction and recursion
Math 3121 Abstract Algebra I Section 0: Sets. The axiomatic approach to Mathematics The notion of definition - from the text: "It is impossible to define.
Section 5.3. Section Summary Recursively Defined Functions Recursively Defined Sets and Structures Structural Induction.
Thomson, “Parthood and Identity Across Time” Thomson’s aim is to argue against the thesis of temporal parts. I Make a Tinkertoy house, H, and place it.
Chapter 3 – Set Theory  .
INM175 Topic 7 1 Module INM175 Discrete Mathematics Topic 7 Set Theoretic Models.
1 Copyright, 1996 © Dale Carnegie & Associates, Inc. Chapter 2: Logic & Incidence Geometry Back To the Very Basic Fundamentals.
Formal Methods in Software Engineering 1
Language: Set of Strings
Copyright © Zeph Grunschlag, Induction Zeph Grunschlag.
Naïve Set Theory. Basic Definitions Naïve set theory is the non-axiomatic treatment of set theory. In the axiomatic treatment, which we will only allude.
ICS 253: Discrete Structures I Induction and Recursion King Fahd University of Petroleum & Minerals Information & Computer Science Department.
Mathematical Preliminaries
Sets and Subsets Set A set is a collection of well-defined objects (elements/members). The elements of the set are said to belong to (or be contained in)
CS 103 Discrete Structures Lecture 13 Induction and Recursion (1)
CS 285- Discrete Mathematics Lecture 4. Section 1.3 Predicate logic Predicate logic is an extension of propositional logic that permits concisely reasoning.
Rosen 1.6, 1.7. Basic Definitions Set - Collection of objects, usually denoted by capital letter Member, element - Object in a set, usually denoted by.
Foundations of Discrete Mathematics Chapters 5 By Dr. Dalia M. Gil, Ph.D.
Mathematical Induction Section 5.1. Climbing an Infinite Ladder Suppose we have an infinite ladder: 1.We can reach the first rung of the ladder. 2.If.
Copyright © Zeph Grunschlag, Induction Zeph Grunschlag.
CompSci 102 Discrete Math for Computer Science March 13, 2012 Prof. Rodger Slides modified from Rosen.
Chapter 5 With Question/Answer Animations 1. Chapter Summary Mathematical Induction - Sec 5.1 Strong Induction and Well-Ordering - Sec 5.2 Lecture 18.
1 Interactive Computer Theorem Proving CS294-9 October 19, 2006 Adam Chlipala UC Berkeley Lecture 9: Beyond Primitive Recursion.
Relations Chapter 9 Copyright © McGraw-Hill Education. All rights reserved. No reproduction or distribution without the prior written consent of McGraw-Hill.
Relations and Their Properties
Dr. Ameria Eldosoky Discrete mathematics
Discrete Mathematics Lecture 6
Normalization.
Chapter 2 Sets and Functions.
Copyright © Cengage Learning. All rights reserved.
Formal methods: Lecture
BFO 2.0 Modularization and Verification
From Classical Proof Theory to P vs. NP
Introduction to Logic for Artificial Intelligence Lecture 2
Relations Chapter 9.
(State) Model-Based Approaches I Software Specification Lecture 35
What is Probability? Quantification of uncertainty.
PROGRAMMING IN HASKELL
PROGRAMMING IN HASKELL
Chapter 5 Induction and Recursion
Discrete Structure II: Introduction
4. The Postulates of Quantum Mechanics 4A. Revisiting Representations
Chapter 2 Boolean Algebra and Logic Gate
Basis and Dimension Basis Dimension Vector Spaces and Linear Systems
Mathematical Background 1
Mathematical Background 1
Copyright © Cengage Learning. All rights reserved.
MA/CSSE 474 More Math Review Theory of Computation
Computer Security: Art and Science, 2nd Edition
Negations of quantifiers
An Ideal Euclidean Model
PROGRAMMING IN HASKELL
REVISION Relation. REVISION Relation Introduction to Relations and Functions.
Formal Program Specification
CS589 Principles of DB Systems Fall 2008 Lecture 4e: Logic (Model-theoretic view of a DB) Lois Delcambre
Presentation transcript:

Formal Theories for Reasoning about Parthood, Connection, and Location Maureen Donnelly University at Buffalo

Outline of Topics Classical Mereology Weaker Mereologies Temporal Mereologies Mereotopologies Location and Containment Relations

Classical Mereology

Formal Theories A formal theory is specified by a collection of formulas of a formal deductive system. These formulas are the axioms of the theory. Think of the axioms as postulates of the theory—the basic assumptions made for reasoning about a given topic. For this class, we will generally assume that our formal deductive system is standard first-order predicate logic with designated symbols for mereological (parthood) relations.

Primitive vs. Defined Relations All formal mereologies assume one or more primitive relations. These relations are not defined. Instead, their logical properties are fixed by the axioms of the theory. All primitives must be interpreted in a way that preserves these logical properties. All other relations in the mereology (the defined relations) are defined in terms of the primitive relations.

Axioms, Definitions, and Theorems In applying a theory, three different kinds of formulas are important: The axioms of a theory are the formulas which fix the logical properties of the primitive relations. The definitions of a theory are the formulas which introduce the defined relations. The theorems of a theory are the formulas that can be derived from the axioms and definitions using the machinery of the background deductive system (e.g. a set of axioms and deductive rules for standard first-order logic).

Models for Formal Theories A model is given for a formal theory by specifying: 1. A domain for the model (i.e., a set of individuals over which the quantifiers range) 2. Interpretations for the primitives of the theory as relations over the domain which satisfy the theory’s axioms (i.e. interpretations of the primitives over the domain which make the axioms true).

Classical Mereology Our version of classical mereology assumes: --2 kinds of entities in the domain of quantification: individuals (x, y, z,…) and collections of individuals (A, B, C,…). --one mereological primitive: the binary relation P where on the intended interpretation Pxy means: x is part of y

Classical Mereology: Definitions PPxy =: Pxy & x  y (x is a proper part of y: x is part of y and x isn’t identical to y) Example: my hand is a proper part of my body Oxy =: z (Pzx & Pzy) (x overlaps y: some object z is part of both x and y) Example: my vertebral column overlaps my pelvis DSxy =: ~ Oxy (x and y are discrete: x and y do not overlap) Example: my nose and my left foot are discrete SUM(x, A) =: y (y  A  Pyx) & z ( Pzx  y (y  A & Ozy)) (x is the sum of the members of A: every member of A is part of x and every part of x overlaps some member of A) nnnn

Classical Mereology: Axioms (A1) Pxx (reflexivity: for any x, x is part of itself) (A2) Pxy & Pyz  Pxz (transitivity: if x is part of y and y is part of z, then x is part of z) (A3) Pxy & Pyx  x = y (antisymmetry: if x is part of y and y is part of x, then x and y are identical)

Classical Mereology: Axioms (A4) ~Pxy  z( Pzx & DSzy) (the strong supplementation principle: if x is not part of y, then there is some z that is part of x and discrete from y) (A5) y y  A  x SUM(x, A) (Universal Fusion Principle: every non-empty collection of individuals has a sum)

Some Theorems of Classical Mereology (T1) PPxy  z( Pzy & DSzx) (Weak Supplementation Principle—if x is a proper part of y, then y has a part that is discrete from x) (T2) z (Ozx  Ozy)  Pxy (if everything that overlaps x also overlaps y, then x is part of y) (T3) z (Ozx  Ozy)  x =y (if x and y overlap the same things, then x and y are identical) (T4) SUM(x, A) & SUM(y, A)  x =y (every collection of individuals has at most one sum) Exercise: Prove that (T1)-(T4) are theorems of Classical Mereology

A Class of Models for Classical Mereology Let S be any non-empty set. Let SB(S) be the set of all non-empty subsets of S. Take SB(S) as the domain of the model. Interpret P as the subset relation on SB(S). Exercise: (i) what interpretations will PP, O, DS, and SUM have in these models? (ii) show that axioms (A1)-(A5) are satisfied over these models.

A Simple Model Let S = {1, 2}. List the members of SB(S) and the interpretations of P, PP, O, DS, and SUM. Verify that the classical axioms are satisfied.

A Spatial Model Let S = 3. SB(3) is the set of all non-empty subsets of 3. P is interpreted as the subset relation on SB(3), PP is interpreted as the proper subset relation on SB(3), O is interpreted as the relation which holds between subsets with a non-empty intersection, and DS is interpreted as the relation which holds between subsets with an empty intersection.

II. Weaker Mereologies

Some Possibly Unintuitive Commitments of Classical Mereology y y  A  x SUM(x, A) For every non-empty collection of objects, there is some object which they all compose. SUM(x, A) & SUM(y, A)  x =y Every collection of objects composes at most one object.

Weaker Mereologies Can be generated by removing axioms from classical mereology: (A1) Pxx (A2) Pxy & Pyz  Pxz (A3) Pxy & Pyx  x = y (A4) ~Pxy  z( Pzx & DSzy) OR:

OR potentially problematic axioms can be replaced by weaker assumptions Define Uyz =: w (Pyw & Pzw) (y and z underlap: there is some object w of which y and z are both parts) (A1) Pxx (A2) Pxy & Pyz  Pxz (A3) Pxy & Pyx  x = y (A4) ~Pxy  z( Pzx & DSzy) (A5*) y y  A & yz(y,z  A Uyz)  x SUM(x, A) (every non-empty collection of pairwise underlapping objects has a sum) Exercise: Prove that the theory consisting of (A1)-(A5*) is weaker than than classsical mereology.

Another weaker mereology Define zDIFyx =:w (Owz  v (Pvx & DSvy & Ovw)) (z is the difference of y in x: z consists of all of x that is discrete from y) (A1) Pxx (A2) Pxy & Pyz  Pxz (A3) Pxy & Pyx  x = y (A4/5*) ~Pxy  z zDIFyx (if x is not part of y, then there is an object which is the difference of y in x) Exercise: Prove that, given (A1)-(A3), (A4/5*) is stronger than (A4) and weaker than (A5).

Exercise Which (possibly) undesirable results do the two preceding mereologies avoid? What might we do to avoid other (possibly) undesirable commitments?

III. Temporal Mereologies

A temporal version of classical mereology (TCM) TCM has objects, collections of objects, and times as disjoint sorts of entities. Use variables: w, x,y,z for objects; A,B, C, for collections; and s,t for times. TCM has one mereological primitive: the ternary relation P (parthood). P takes two objects and a time as its arguments. Pxyt Is intended as: Object x is part of object y at time t.

Defined Relations PPxyt =: Pxyt & x  y (x is a proper part of y at t: x is part of y at t and y is not part of x at t ) Oxyt =: z (Pzxt & Pzyt) (x and y overlap at t: some object z is part of both x and y at t) DSxyt =: ~Oxyt (x and y are discrete at t: x and y do not overlap at t) Ext =: Pxxt (x exists at time t: x is part of itself at t)

TCM’s First Five Axioms (AT1) t Ext (every object exists at some time) (AT2) Pxyt  (Ext & Eyt) (if x is part of y at t, then both x and y exist at t) (AT3) (Pxyt & Pyzt)  Pxzt (if x is part of y at t and y is part of z at t, then x is part of z at t) (AT4) (Pxyt & Pyxt)  x = y (if x is part of y at t and y is part of x at t, then x and y are identical) (AT5) (Ext & ~Pxyt)  z(Pzxt & DSzyt) (if x exists at t and x is not part of y at t, then x has some part z at t that is discrete from y at t)

Temporal Relations for Collections FP(A, t) =: x(x  A) & x (x  A  Ext) (A is fully-present at t: there is some member of A and all members of A exist at t) SP(A, t) =: x (x  A & Ext) (A is some-present at t: some member of A exists at t) NP(A, t) =: x ( x  A  ~Ext) (A is non-present at t: no members of A exist at t)

Summation Relations Time-Dependent General Summation: SUM(z, A, t) =: Ezt & y (Oyzt  x(x  A & Oyxt)) (z is a sum of the A’s at t) SM1(x, A) =: t (Ext  SUM(x, A, t)) (whenever x exists, x is a sum of A’s) SM2(x, A) =: t (SP(A, t)  SUM(x, A, t)) (whenever A is some-present, x is a sum of A’s) SM3(x, A) =: t (FP(A, t)  SUM(x, A, t)) (whenever A is fully-present, x is a sum of A’s) SM4(x, A) =: t ((Ext  SP(A, t))  SUM(x, A, t)) (whenever x exists or A is some-present, x is a sum of A’s) Exercise: Give examples of objects in the world that seem to stand in the summation relations above.

A Summation Axiom for TCM (AT6) SP(A, t)  z SUM(z, A, t) (if some member of A exists at t, then there is a sum of A’s at t) NOTE: We could have instead used one of the time-independent relations to generate a different temporal version of CM. Exercise: Write out each alternative summation axioms for each of the time-independent summation relations. Determine whether each of these alternative summation axioms is stronger or weaker than (AT6).

A Class of Models for TCM Let S be any non-empty set. Let T be a non-empty set of times. Let (S) be the set of subsets of S. Let OB be any set of functions from T into (S) s.t. for any f, g  OB and any A  OB: there is some t, s.t. f(t)   if f(t) = g(t), then f = g if f(t)   and f(t)  g(t), then there is some h  OB s.t. h(t)  f(t) and h(t)  g(t) =  if there is some h  A s.t. h(t)  , then there is some j  OB s.t. j(t) = h  A h(t) if f(t)  g(t)  , then there is some h  OB s.t   h(t)  f(t)  g(t)

A Class of Models (con’t) If we interpret P  {<f, g, t>: f, g  OB, t  T, f(t)  , & f(t)  g(t) } then (AT1)-(AT6) are satisfied. ( Exercise: verify!!) Exercise: What are the interpretations of the defined relations?

IV. Mereotopologies

Connection We can add to any atemporal mereology the binary connection predicate C where Cxy is interpreted as: x is connected to y (i.e. x is zero distance from y) (For a temporal theory, we would use a ternary predicate and Cxyt would be interpreted as: x is connected to y at time t.)

Axioms for Connection (weaker version ) (AC1) Cxx (connection is reflexive) (AC2) Cxy  Cyx (connection is symmetric) (AC3) Pxy  z(Czx  Czy) (if x is part of y, then every thing that is connected to x is also connected to y)

Additional Defined Relations ECxy =: Cxy & DSxy (x and y are externally connected: x and y are connected but do not overlap) TPxy =: Pxy & z(ECzx & ECzy) (x is a tangential part of y: x is a part of y that is externally connected to something which is externally connected to y) IPxy =: Pxy & ~ TPxy (x is an interior part of y: x is part of y, but not a tangential part of y)

Additional Defined Relations SCx =:yz SUM(x, {y, z})  Cyz x is self-connected: any two parts that make up all of x are connected to each other (i.e. there is no way of dividing x into disconnected parts)

Comparing Mereotopologies A weaker mereotopology MT- uses 2 mereotopological primitives (P and C) and 8 axioms: (A1) Pxx (A2) Pxy & Pyz  Pxz (A3) Pxy & Pyx  x = y (A4) ~Pxy  z( Pzx & DSzy) (A5) y y  A  x SUM(x, A) (AC1) Cxx (AC2) Cxy  Cyx (AC3) Pxy  z(Czx  Czy)

A Stronger Mereotopology MT+ has C (connection) as its only mereotopological primitive and defines: Pxy =: z(Czx  Czy) (x is part of y: everything that is connected to x is also connected to y) 5 Axioms: (AC1) Cxx (AC2) Cxy  Cyx (A3) Pxy & Pyx  x = y (A4) ~Pxy  z( Pzx & DSzy) (A5) y y  A  x SUM(x, A) Exercise: Verify that MT+ is stronger than MT.

Models Let SB(3) is the set of all non-empty subsets of 3 with P interpreted as the subset relation and C interpreted as the relation that hold between two subsets when they are zero-distance apart…i.e.when the closures of the subsets intersect. Exercise Show that this structure satisfies all of the axioms of MT-, but not those of MT+. Exercise Find a model of MT+.

(dis)advantages of the Stronger Mereotopology By defining parthood in terms of connection as Pxy =: z(Czx  Czy) We may rule out cases in which an object has a point-sized, line-sized, or surface-sized gap in the middle. On the other hand, some ontologists want to rule out these sorts of irregularly-shaped objects anyway (along with lower-dimensional objects). If so, the stronger mereotopology may be preferable because it is more economical.

V. Location and Containment Relations

Objects and Regions We can expand a basic mereology by adding vocabulary for describing where objects are positioned in a fixed background space. To do this, we need to expand our domain of quantification to include immaterial spatial regions as well as the material objects (and immaterial holes or cavities) which are located at those regions. To the formal vocabulary, we add the unary region function r which takes object or region terms as arguments and where r(x) is interpreted as the spatial region at which x is exactly located.

Axioms for the Region Function (AR1) PPxy  PPr(x)r(y) (if x is a proper part of y, then x’s region is proper part of y’s region) (AR2) r(r(x)) = r(x) (x’s spatial region is its own spatial region)

Region Containment RCONxy =:Pr(x)r(y) x is region-contained in y: x’s region is part of y’s region. Examples: My heart is region-contained in my middle mediastinal space. My larnyx is region-contained in my neck. NOTE: The larnyx is part of the neck. But the heart is not part of the middle mediastinal space. y x

Convex Hulls The convex hull of an object or region x is the smallest convex region in which x is region-contained. We may add a separate primitive function mapping each object or region to its convex hull. ch(x) is x’s convex hull.

Axioms for the convex hull function (AR3) Pr(x)ch(x) (x’s region is part of x’s convex hull) (AR4) RCONxy  Pch(x)ch(y) (if x is r-contained in y, then x’s convex hull is part of y’s convex hull) (AR5) ch(ch(x)) = ch(x) (x’s convex hull is its own convex hull)

Surrounding SURxy =: Pr(x)ch(y) & DSr(x)r(y) x is surrounded by y: x’s region is part of y’s convex hull but x’s region does not overlap y’s region. Examples: My pleural space is surrounded by my pleural membrane. A bolus of food in my stomach is surrounded by the wall of my stomach. y x

Partial Containment PCONxy =: Or(x)ch(y) x is partially contained in y: x’s region overlaps y’s convex hull. Examples: My esophagus is partially contained in my thoracic cavity. My tooth is partially contained in its socket. y x

Logical Properties-- Transitivity RCON is transitive. SUR and PCON are NOT transitive. Examples: My heart is region-contained in my middle mediastinal space and my middle mediastinal space is region-contained in my thoracic cavity. So my heart is region-contained in my thoracic cavity. A filling is partially contained in my tooth and my tooth is partially contained in its socket, but the filling is NOT partially contained in the socket.

Logical Properties-- Interaction with Parthood The three defined containment relations interact differently with the parthood relation. For example: If my heart is region-contained in my middle mediastinal space then any part of my heart is also region-contained in my middle mediastinal space. BUT: My tooth is partially contained its socket even though some are its parts (e.g. its crown) are NOT partially contained in the socket.

Exercise Derive as many theorems as you can for the relations RCON, SUR, and PCON and in their interaction with each other and with the merological relations P and O.