Lecture 11: Safra s Determinization Construction

Similar documents
Addition & Subtraction of Polynomials

Flow Networks Alon Efrat Slides courtesy of Charles Leiserson with small changes by Carola Wenk. Flow networks. Flow networks CS 445

f(x) dx with An integral having either an infinite limit of integration or an unbounded integrand is called improper. Here are two examples dx x x 2

8.5 Circles and Lengths of Segments

Reinforcement learning

Primal and Weakly Primal Sub Semi Modules

Chapter 7. Kleene s Theorem. 7.1 Kleene s Theorem. The following theorem is the most important and fundamental result in the theory of FA s:

Bipartite Matching. Matching. Bipartite Matching. Maxflow Formulation

() t. () t r () t or v. ( t) () () ( ) = ( ) or ( ) () () () t or dv () () Section 10.4 Motion in Space: Velocity and Acceleration

EE Control Systems LECTURE 8

LAPLACE TRANSFORMS. 1. Basic transforms

4.8 Improper Integrals

1. Find a basis for the row space of each of the following matrices. Your basis should consist of rows of the original matrix.

Faraday s Law. To be able to find. motional emf transformer and motional emf. Motional emf

ME 141. Engineering Mechanics

ISSUES RELATED WITH ARMA (P,Q) PROCESS. Salah H. Abid AL-Mustansirya University - College Of Education Department of Mathematics (IRAQ / BAGHDAD)

the king's singers And So It Goes the colour of song Words and Vusic by By Joel LEONARD Arranged by Bob Chilcott

Dividing Algebraic Fractions

Classification of Equations Characteristics

CSC 373: Algorithm Design and Analysis Lecture 9

Language Processors F29LP2, Lecture 5

Circuits 24/08/2010. Question. Question. Practice Questions QV CV. Review Formula s RC R R R V IR ... Charging P IV I R ... E Pt.

Partial Fraction Expansion

BINOMIAL THEOREM OBJECTIVE PROBLEMS in the expansion of ( 3 +kx ) are equal. Then k =

A study Of Salt-Finger Convection In a Nonlinear Magneto-Fluid Overlying a Porous Layer Affected By Rotation

Physics 15 Second Hour Exam

Sections 3.1 and 3.4 Exponential Functions (Growth and Decay)

Science Advertisement Intergovernmental Panel on Climate Change: The Physical Science Basis 2/3/2007 Physics 253

Laplace Examples, Inverse, Rational Form

graph of unit step function t

OH BOY! Story. N a r r a t iv e a n d o bj e c t s th ea t e r Fo r a l l a g e s, fr o m th e a ge of 9

_ J.. C C A 551NED. - n R ' ' t i :. t ; . b c c : : I I .., I AS IEC. r '2 5? 9

Bounding the Locality of Distributed Routing Algorithms

Physics 201, Lecture 5

Calculus 241, section 12.2 Limits/Continuity & 12.3 Derivatives/Integrals notes by Tim Pilachowski r r r =, with a domain of real ( )

Consider a Binary antipodal system which produces data of δ (t)

The Area of a Triangle

International Journal of Mathematical Archive-5(6), 2014, Available online through ISSN

Available online Journal of Scientific and Engineering Research, 2018, 5(10): Research Article

can be viewed as a generalized product, and one for which the product of f and g. That is, does

156 There are 9 books stacked on a shelf. The thickness of each book is either 1 inch or 2

EECE 301 Signals & Systems Prof. Mark Fowler

T h e C S E T I P r o j e c t

Chapter 2. Kinematics in One Dimension. Kinematics deals with the concepts that are needed to describe motion.

Satellite Orbits. Orbital Mechanics. Circular Satellite Orbits

P a g e 5 1 of R e p o r t P B 4 / 0 9

0 for t < 0 1 for t > 0

Applications of these ideas. CS514: Intermediate Course in Operating Systems. Problem: Pictorial version. 2PC is a good match! Issues?

D zone schemes

Today - Lecture 13. Today s lecture continue with rotations, torque, Note that chapters 11, 12, 13 all involve rotations

A Structural Approach to the Enforcement of Language and Disjunctive Constraints

MTH 146 Class 11 Notes

Some algorthim for solving system of linear volterra integral equation of second kind by using MATLAB 7 ALAN JALAL ABD ALKADER

Exponential Sawtooth

Molecular Evolution and Phylogeny. Based on: Durbin et al Chapter 8

The sphere of radius a has the geographical form. r (,)=(acoscos,acossin,asin) T =(p(u)cos v, p(u)sin v,q(u) ) T.

Maximum likelihood estimate of phylogeny. BIOL 495S/ CS 490B/ MATH 490B/ STAT 490B Introduction to Bioinformatics April 24, 2002

A L A BA M A L A W R E V IE W

On Fractional Operational Calculus pertaining to the product of H- functions

Motion on a Curve and Curvature

C o r p o r a t e l i f e i n A n c i e n t I n d i a e x p r e s s e d i t s e l f

Solutions to assignment 3

Homework 5 for BST 631: Statistical Theory I Solutions, 09/21/2006

General Non-Arbitrage Model. I. Partial Differential Equation for Pricing A. Traded Underlying Security

Minimum Squared Error

Computer Aided Geometric Design

Minimum Squared Error

ARC 202L. Not e s : I n s t r u c t o r s : D e J a r n e t t, L i n, O r t e n b e r g, P a n g, P r i t c h a r d - S c h m i t z b e r g e r

10 Statistical Distributions Solutions

Ans: In the rectangular loop with the assigned direction for i2: di L dt , (1) where (2) a) At t = 0, i1(t) = I1U(t) is applied and (1) becomes

PHYSICS 1210 Exam 1 University of Wyoming 14 February points

e t dt e t dt = lim e t dt T (1 e T ) = 1

Technical Vibration - text 2 - forced vibration, rotational vibration

Data Structures. Element Uniqueness Problem. Hash Tables. Example. Hash Tables. Dana Shapira. 19 x 1. ) h(x 4. ) h(x 2. ) h(x 3. h(x 1. x 4. x 2.

The Nehari Manifold for a Class of Elliptic Equations of P-laplacian Type. S. Khademloo and H. Mohammadnia. afrouzi

( ) ( ) ( ) ( ) ( ) ( y )

Ch.4 Motion in 2D. Ch.4 Motion in 2D

CONTRIBUTIONS TO THE THEORETICAL STUDY OF THE PRECISION SOWING MACHINES DYNAMICS

Algorithmic Discrete Mathematics 6. Exercise Sheet

Bisimulation, Games & Hennessy Milner logic p.1/32

Chapter Finite Difference Method for Ordinary Differential Equations

ENGR 1990 Engineering Mathematics The Integral of a Function as a Function

Week 10: DTMC Applications Ranking Web Pages & Slotted ALOHA. Network Performance 10-1

Chapter Introduction. 2. Linear Combinations [4.1]

Math 4318 : Real Analysis II Mid-Term Exam 1 14 February 2013

Previously. Extensions to backstepping controller designs. Tracking using backstepping Suppose we consider the general system

Properties of Logarithms. Solving Exponential and Logarithmic Equations. Properties of Logarithms. Properties of Logarithms. ( x)

Lecture 22 Electromagnetic Waves

Week 8. Topic 2 Properties of Logarithms

Price Discrimination

PHYSICS 102. Intro PHYSICS-ELECTROMAGNETISM

Introduction to SLE Lecture Notes

Motion. ( (3 dim) ( (1 dim) dt. Equations of Motion (Constant Acceleration) Newton s Law and Weight. Magnitude of the Frictional Force

X-Ray Notes, Part III

Transformations. Ordered set of numbers: (1,2,3,4) Example: (x,y,z) coordinates of pt in space. Vectors

The solution is often represented as a vector: 2xI + 4X2 + 2X3 + 4X4 + 2X5 = 4 2xI + 4X2 + 3X3 + 3X4 + 3X5 = 4. 3xI + 6X2 + 6X3 + 3X4 + 6X5 = 6.

Some basic notation and terminology. Deterministic Finite Automata. COMP218: Decision, Computation and Language Note 1

The Covenant Renewed. Family Journal Page. creation; He tells us in the Bible.)

Electric Potential. and Equipotentials

MEEN 617 Handout #11 MODAL ANALYSIS OF MDOF Systems with VISCOUS DAMPING

Transcription:

Lece 11: Sf Deeminizion Concion In hi lece we hll ee concion de o S. Sf h n nondeeminiic Büchi omon of ize n ino n eivlen deeminiic Rin omon of ize 2 O(nlogn). A n immedie coolly we ee h we cn comlemen nondeeminiic Büchi omon ino deeminiic See omon of ize 2 O(nlogn). Sf concion i one of he meiece of hi e. Thee e nme of icle h y o demyify nd exlin he concion [2, 4]. They y nd exlin why imle ide do no wok nd led yo o Sf concion. A he ik of ofcing me I hll no ke hi oe nd ined y o ive Sf concion fom n nlyi of he n-ee of nondeeminiic Büchi om. The em hee i o exlin he ecil ce of he e ce conced omehing h ie nlly fom he ce of cceing n (he hn comliced ce oined y eence of efinemen of imle ide h do no wok). A nondeeminiic omon h mny n on n in wod. The deeminiic omon m imle ll hoe n (o ll he elevn one) in ingle one. The e of ll n of NBA A on wod w wold e n infinie ee in which ech h i n of A on w. (Thi i imil o he n of nivel lening omon, i.e. one in which ll he e e e. Of coe, he ccence cieion i diffeen. We only demnd h le one h hogh hi ee i cceing.) Hee i NBA A f nd i n on he wod...,,, We cold deeminiiclly imle hi omon level y level. Th i ey mch ll h we cn do in deeminiic imlion. B he nme of node in level i no onded. So, we need oxime nd minin ome onded mon of infomion. In he o clled owe-e concion, we oxime level y he e of e h lel ome veex he level. Th, we hve n omon of ize 2 n. Nex, we need o decide on he ccence condiion. Demnding h h infiniely mny level conin finl e i oo geneo. In he ove n, he cceing e F e in evey level in he n ee, howeve hee i no h in hi ee h vii F infiniely ofen (nd hence... i no cceed y he ove omon). Th he owee omon A = (2 Q, Σ,δ, {}, {X X F }) wih δ (X,) = { X. δ(,)} 1

cce moe wod hn A. The owe-e omon conced fom A f cce... hogh A doe no. Hee i n cceing n: Fo ny wo e X,Y of e fom Q nd wod, le wie X g Y if fo ech Y hee i X ch h hee i n fom o on h vii ome e in F. Fo exmle, in he omon A f, {,,} g {,}. We now decie diffeen ccence cieion fo he owee omon. Le S 1 0 S 2 2... e he n of he owee omon on 1 2... We cce ovided we cn x decomoe hi n S 1 0 2 Si1 Si2... ch h fo ech S ik+1, hee i S ik ch h x k+1 g. We clim h ccence nde hi cieion gnee h A cce w. We cn ee hi follow: Conc DAG whoe node level k e he e in S ik. An edge i dwn fom node level i o node level i + 1 if nd only if x i+1 g. Lel hi edge y ch good n fom o. Thi finiely nching DAG h infiniely mny node echle fom he veice level 1. Th, hee m e n infinie h hogh hi DAG. The lel long he edge of hi h give n cceing n fo A on w. Cn we nle hi ccence condiion Büchi ccence condiion (y lowing he e ce if necey?) The ide i imil o he one ed o nle ABA ino NBA. Le A m = (Q m, Σ,δ m, ({}, ), {(X,X) X Q} whee Q m = {(X,Y ) Y X Q} δ m ((X,Y ),) = (δ (X,),δ (Y,) δ(x,) F) δ m ((X,X),) = (δ (X,),δ (X,) F) if X Y We imle he owee omon one level. In he ohe level, we kee ck of he e of e h hve een eched vi h h viied F. When hee wo e e he me, we ee he econd e nd oceed. Th we hve deeminiic omon A m wih L(A m ) L(A). In [2], he omon A m i clled he mked owee omon. Qie clely, L(A m ) will no in genel e he me L(A). Fo innce, he mked owee omon conced fom A f doe no cce... which i cceed y A f. 2

In genel, if hee i ded e d( noncceing e wih elf-loo on ll lee which i fhe no echle fom ny cceing e) i echle vi ome efix of w hen A m will no cce w. Thi i ece he e d e in evey level of he nee eyond finie efix on evey wod nd hee i NO h in he omon o ech d vi n cceing e. How do we ge ond hi? The nive ide wold e omehow modify he concion of A m o do ome elemen of Q fom he cen e (o h ny ded-e e doed.) B h i wh nondeeminiic omon doe! I imly do ll he e exce he one h e on he cceing h. So we need ee ide. I n o h wh we need o do i o he omon L(A m ) no igh he eginning, ined i fe ome iniil efix h een ed. In ome ene, we wi fo ll he ele h o nch off nd n he A m on coe of he nee. Of coe, we hve o do ll hi deeminiiclly. Conide he following fige. I decie nee of n NBA A on wod w = 1 2... which i cceed y A. In he fige, we hve mked one cceing h in hi nee, nd indiced he e of F h e long hi h y dkened cicle. S 3 S 1 3 1 0 2 S 0 S 2 y x 0 y 0 x 1 y 1 x 2 y 2 x 3 Hee 0, 1,... e he e fom F eched long he mked cceing n. We hve lced lice S 0,S 1,... on he -nee ooed 0, 1,... S 0 i he e of node eched 3

fom 0 on eding he egmen x 0, S 1 i he e of node eched fom 1 on eding he egmen x 1 nd o on. w = yx 0 y 0 x 1 y 1... (The egmen y i i he ffix oined when we do x i fom he wod h ke i o i+1.) y i x i+1 Noice h S i g S i+1. Thi migh led o elieve h we imly hve o he omon A m 0. Unfonely, S 1 need no e e of e viied y he omon A ed e 0 (on x 0 y 0 x 1 ), i.e. δ ( 0,x 0 y 0 x 1 ) cold e ic ee of S 1 = δ ( 1,x 1 ). A indiced in he fige elow, he e my e in δ ( 0,x 0 y 0 x 1 ) no in S 1. S 1 0 S 0 1 B hen i δ ( 1,x 1 y 1 x 2 ) he me δ ( 2,x 2 ) (i.e. cn we he A m omon 2?) Well, once gin δ ( 1,x 1 y 1 x 2 ) cold e ic ee of δ( 2,x 2 ). Noe h if δ ( i,x i y i ) = δ ( i+1,) fo ome hen δ ( i,x i y i v) = δ ( i+1,v) fo ll exenion v. i i+1 Now, le y h i nd i+1 mege if hee i n ch h δ ( i,x i y i ) = δ ( i+1,). We cn eily exend hi noion of meging o i nd j fo ny i of occnce of cceing e (long he cceing h nde conideion.) If i mege wih i+1 nd i+1 mege wih i+2 hen clely i nd i+2 lo mege. I i ey o check h mege i n eivlence elion h divide 0, 1... ino egmen ( 0,... i ), ( i+1... j ),... Alo noe h if i nd i+1 do no mege, hen fo ny exenion of x i y i, δ ( i,x i y i ) i ic ee of δ ( i+1,). B he e δ( i,) i e of Q nd hence finie. Th, mege i n eivlence elion of finie index. Th hee i n eivlence cl of mege coniing of { j j N 0 } fo ome N 0. 0 1 i N 0 w Soe, h w = w whee N0, hen we clim h he mked e omon ed e F lelling N0 cce he wod w. Th i, he omon A m = 4

(2 Q 2 Q, Σ,δ m, ({}, ), {(X,X) X Q} whee δ m i efoe, cce he wod w. Le (X 1,X 2 ) e he e eched fe ome efix v of w. Hence hee i ome j n h gone j no j+1. The e N0 nd j+1 m mege ome efix w 1 of w. A h oin he e of A m wold e (Y,Y ) fo ome Y. Th, he omon A m vii he cceing e infiniely ofen on w. Th, o deeminiic omon m (deeminiiclly!) ge h in he nee nd F lelled oiion in hi h (coeonding o P N0 ) nd n he mked e omon (which i deeminiic omon) A m fom hee on. Th, in mnne ie eminicen of he oof of McNghon heoem, we hve hown h nondeeminim i needed only in finie efix of he n. The eion i how o ge id of he iniil nondeeminim. A in he oof of McNghon heoem, he ide wold e o n ll oile coie ined of geing which coy o n nd mege coie wheneve oile. Th i, ech oin in he n if he wod ed o f led o ome finl e we fok off coy of A m hi oin. We lo mege coie h e he me e. Le exmine he coeondence wih he concion in Lece 8 cloely: 1. The imle owee omon A will ly he ole of A U. I ole i o idenify he efixe which coie of he deeminiic Büchi om e o e foked off. 2. The ole of he deeminiic Büchi omon A V will e lyed y collecion of deeminiic Büchi om {A m F }. Afe eding finie wod w, if he omon A h eched he e X hen, fo ech X F coy of A m i o e foked off. 3. Oeve h if hen A m nd A m diffe only in he e. Th i, he e ce of ll he A m i h of A m. Th he oeion of meging diffeen coie ill mke ene. In icl, if wo diffeen coie e he me e we mege hem ogehe. 4. The nme of e in A m i 2 2n. A ech e, we cold fok o F coie (deending on which e in F e in he cen e). To ene h if coy lo j mege (wih lowe nmeed coy) down hen lo j i emy (i.e. i cie he e ) in he nex e, i ffice o kee 2 2n + F lo. Thi ene h ny lo h ecome vcn (ece i h meged wih ome ohe coy) y vcn le fo one move. 5. A n i cceing eciely when hee i ome lo j which i only finiely ofen (o h he coy j imle ome A m on ome ffix of w) nd omon lo j vii he finl e infiniely ofen (i.e. he e in he jh lo i of he fom (X,X) infiniely ofen). Thi led o he following concion of he (dole exonenil ized) deeminiic 5

omon A d cceing L(A). Le K = 2 2n + F. Q d d = 2 Q (2 Q 2 Q { }) K = ({},,,... ) Le (S,W 1,W 2,...W K ) e e. To come δ d ((S,W 1,W 2...,W K ),), fi come he le V = (δ (S,),δ m (W 1,),δ m (W 2,)...δ m (W K,)) wih he ndending h δ m (,) =. Soe { 1, 2,..., l } = δ(s,) F. We need o fok coy of A i m fo ech i. Pick he l lowe nmeed lo in V h e occied y nd elce hem y ({ 1 }, ), ({ 2 }, )...({ l }, ) eecively. Now, if wo o moe lo e occied y he me e of A m, hen elce ll he lowe nmeed ch lo y. Thi V i δ d ((S,W 1,W 2,...W K ),). Finlly, he ccence condiion i Rin ccence condiion coniing of K i (F i,i i ). F i conin ll he le whee he ih coodine i. I i conin ll he le whee he ih coodine i n cceing e of A m, h i i i of he fom (X,X) fo ome X. I i ie ey o check, ing he me ide in he oof in Lece 8, h hi omon cce wod eciely when i i cceed y L(A). How o nfom hi dole exonenil ized omon o ingle exonenil ized omon i he oic of he nex ecion. 1 Sf Concion Sf concion incooe he ce of he nee ino he e ce of he deeminiic omon. We hll ille hi ing he following comlex omon which cce he e of ll wod wih infiniely mny :,,,, v, Hee i efix of he nee of hi omon on he in... 6

v v Thee e even occnce of cceing e in he n on he wod. A n when hee e e enconeed coie of A m i foked o. The e eched y hee even coie i mked nex o hee e. Moeove he e eched y comlee owee omon i mked nex o he oo. {,,v,,} ({},{}) ({,},{,}) ({,,v},{,}) ({},{}) ({},{}) v ({},{}) ({},{}) v In he mon A d hi wold hve een eeened y le ({,,v,,}, ({,}, {,}), ({}, ), ({,,v}, {,}), ({}, ), ({}, )) (wih lge nme of ineeed which we hve omied o void cle). Sf concion oe moe infomion. I ecod elionhi eween hee vio coie. Fo exmle, he hee coie coeonding o he level 2, nd he 7

level 3 nd 4 e woking on indeenden ee. Thee e wo coie woking in he ee coeonding o he coy ed w... he level 3 nd o on. I wold ecod hee deendencie nd oe he e ee. ({,,v,,},{,,v,}) ({,},{,}) ({,,v},{,}) ({},{}) ({},{}) ({},{}) ({},{}) ({},{}) I lo e coy of A m ined of A fo he omon h fok o deeminiic om. Thi llow fo nifom emen of ll he node in he ee. (We hve no elimined dlice coie hee. We will ge down o h oon.) In A d we minined he e le nd o he index i llow o efe o icl coy of A m. To do hi hee, we me h hee i n infinie odeed e of node n 1 < n 2... nd ech ime we dd node in he ee, we ick he nex node fom hi eence nd e i. Fo exmle, wih hi he ove ee wold look like: n 1 ({,,v,,},{,,v,}) ({,},{,}) n 3 ({,,v},{,}) n 4 n 2 ({},{}) ({},{}) n 5 n 7 n 6 n 8 ({},{}) ({},{}) ({},{}) Noice h he node e nmeed in he ode in which he node wee inodced ino he ee (We me ome fixed odeing on Q o when mlile childen e dded o node in ingle e, he ode i deemined y he odeing on Q). To ndend he ehvio of hi omon le ee wh hen o hi e when he nex in lee, y i ed: 1. Fi, we imle hi move in ech of he coie of A m (one e node in he ee). 2. Then, we need o inodce new coie of A m fo ech e in f F h i eched hi e. Whee do we inodce hee coie? Noice h ech ch f e long ome e of h ing he oo (ince he e of e lelling en of node i lwy ee of h lelling child). Fo ech ch h, find he highe level ch node on h h nd dd child lelled ({f}, ). In hi exmle, i n o h ll hee highe level node e leve nd o we dd childen o hee leve. B hi need no lwy e he ce. Fo he ove omon hi give: 8

n 1 ({,,v,,},{,,v,}) ({,,v},{,,v}) n 3 ({,,v},{,,v}) n 4 n 2 ({},{}) ({,v},{}) n 5 n 7 n 6 n 8 ({,,v},{}) ({,},{}) ({,,v},{}) ({},{}) n 9 n 10 n 11 n 12 ({},{}) ({},{}) ({},{}) 3. If he e lelling node nd i child e of he fom (X,Y 1 ) nd (X,Y 2 ) hi indice h he coie of A m ed he node nd i child hve meged nd o we imly delee he child nd nfe ll i childen o he en. In he ove exmle, doing hi we ge: n 1 ({,,v,,},{,,v,}) ({,,v},{,,v}) n 3 ({,,v},{,,v}) n 4 n 2 ({},{}) ({,v},{}) n 5 n 7 n 6 n 8 ({,,v},{}) ({,},{}) ({,,v},{}) ({},{}) n 9 n 10 n 11 n 12 ({},{}) ({},{}) ({},{}) Th, nlike in A d we do no cively mege ny i of coie h hve eched he me e only en-child i h hve eched he me e. Noice h hi ene h long ny h in he ee he nme of e in he fi comonen of he node lel decee icly. Th ny h in ch ee i of lengh onded y Q. On he ohe hnd, hee i no ond on he nme of childen of given node. (Oevion 1: Noe h wheneve child of n mege wih n, n m e lelled y e of he fom (X,X), i.e. n cceing e of A m. Thi i ece if node lelled (X,Y ) h childen wih lel (X 1,Y 1 )...(X k,y k ) hen Y = X 1 X 2...X k.) The iniil e of he omon i he ee wih ingle node n 1 wih he e ({}, ). The niion elion decied ove gnee h in ny echle e, if he node n i i he en of n j nd he lel of n i nd n j e (X i,y i ) nd (X j,y j ) eecively, hen 9

X j i ic e of X i. We ke o e of e o e ee ove n 1,n 2... lelled y e of A m h fhe ify hi oey. Th he ee coniing e hve deh onded y Q. Howeve, ince we hve n infinie e of lel he omon we hve conced h infiniely mny e. Finlly, we hve o define he ccence condiion. I cce wod w, if hee i node n i ch h long he nie infinie n on w n i e in ll he e in ome infinie ffix nd moeove n i i lelled y n cceing e of A m infiniely ofen. The coecne of hi concion i ie ovio: I imly minin he e of A d in moe comliced fom (nd wih ome edndncy) nd o i coecne follow fom he coecne of A d. So i eem like we hve ken e in he wong diecion: Fom imle dole exonenil omon o comlex infinie e omon. B one illin oevion of Sf ke ll he wy o 2 O(nlogn) concion. Hi oevion i he following: If Q e in he e lelling mlile h in he ee (i.e. e of he Sf omon), hen we cn do i fom ll one! (By hi we men h if node n i i no long he choen h fo nd i lel i (X,Y ) hen we my elce hi lel y (X \ {},Y \ {})). Thi choice of which h i o kee ck of h o e mde ceflly. Childen of ny node e odeed fom lef o igh in he ode in which hey wee ineed ino he ee. Given wo h ing he oo, he one h move lef he oin of hei nching i id o e o he lef of he ohe. Sf ide i o kee he e in he lef mo h fom he oo (long which e). Wih hi ide, he ove e ecome: n 1 ({,,v,,},{,,v,}) n 2 ({},{}) ({,,v},{,,v}) n 3 ({},{}) n 4 ({,v},{}) n 5 n 10 n 6 ({},{}) n 12 ({},{}) ({},{}) ({},{}) n 9 n 11 ({},{}) We cn delee node lelled ({}, {}) ince hey ly no fhe ole in he n. Th he e eched on wold clly look like: n 1 ({,,v,,},{,,v,}) n 2 ({},{}) ({,,v},{,,v}) ({,v},{}) n 5 n 3 ({},{}) n 9 The modified omon A, h i e of e ee ove he e node n 1,n 2... lelled y e of A m in ch wy fo ny Q, hee i he mo one h (ing 10

he oo) whoe lel conin. The niion elion i comed ove wih he following ddiionl e: 4. If e long lel long mlile h ing he oo, delee i fom ll he lef-mo h. The ccence cieion conine o e he me. Why doe hi modified omon A cce he me lngge A (nd A d )? Evey wod cceed y A i cceed y A. Thi i ece, doing ome e fom he e lelling he ee coeond o nning coie of A m in ch wy h occionlly we do ome e fom he e. Thi doe no incee he fmily of wod cceed hown y he following execie. Execie: Le A m e he omon defined follow: A m = (Q m, Σ,δ m, ({}, ), {(X,X) X Q} whee Q m = {(X,Y ) Y X Q} δ m((x,y ),) = {(δ (X,) \ Z,δ (Y,) δ(x,) F \ Z) Z Q} δ m((x,x),) = {(δ (X,) \ Z,δ (X,) F \ Z Z Q}) if X Y Show h L(A m) L(A m ). Hin: Simly how h if (X,X F) (Z,Z) in hi omon hen fo ech z Z hee i ome x X ch h x g z. So we e lef wih howing h ny wod cceed y A i lo cceed y he modified omon A. Le ρ = 1 0 2... 1 e n cceing n. Le T 1 0 2 T1... e he coeonding n in A. We ocie eence of node N 1,N 2..., N i T i wih he e i of ρ. In T i, e N i o e he node he highe level nme conining i (Thi node need no e lef.) Noice h ech e one of he following hing my hen: 1. N i+1 i child of N i. Thi hen if i+1 F nd he h fom oo N i+1 i he lef-mo h wih i+1. 2. N i+1 i n nceo of N i. Thi hen of N i nd evel of i nceo hve he me lel N i+1 nd hey ll ge meged wih N i+1. 3. N i+1 i eched vi lef jm. Oevion 2: If he ocied node eche node n fom i decenden (y hi we men N i+1 = n nd i i eched ing ce 2 ove) hen n m e lelled y n cceing e of A m. Thi follow fom Oevion 1 nd he condiion fo meging e. Oevion 3: By Oevion 1, we do no need o kee he econd comonen of he 11

lel of he ee hi cn e comed fom he fi comonen of he childen. (Sf concion e hi oimizion.) Fi we oeve h if n 0 e N i fo infiniely mny i hen he node n 0 i lelled y n cceing e of A m infiniely ofen. Thi i ece, ech ime he h vii n j F, he node N j wold move o child of n 0. B i en o n 0 in he fe. B hi hen only when ome child of n 0 mege wih i. B, y Oevion 2, n 0 m e lelled y n cceing e wheneve hi hen. Ohewie, hee i oin j ch h N i n 0 fo ll i j. A he l oin, y N i, when he ocied node moved fom n 0 i m hve moved o child of n 0. Moeove hee e only finiely mny childen of he oo nd in icl only finiely mny o he lef of N i. So wheneve he ocied node en o level 2 (i.e. he level of childen of he oo) hee e only finiely mny choice (i cnno eve ech child of he oo h e o he igh of N i wiho viiing he oo, we hve ledy moved eyond he l vii o he oo). Th, he only wy o ge o ny of hee i eihe y mege fom elow o y lef jm. B lef jm o node level 2 cn hen only finie nme of ime (ince hee e only finiely mny of hee veice h lie o he lef of N i hi level). Th, hee i oin eyond which no veex level 2 i eched y lef jm. Beyond hi oin eihe ome veex level 2 i viied infiniely ofen vi mege fom elow, which (y Oevion 2) gnee h hi node i lelled y cceing e of A m infiniely ofen o ele hee i oin eyond which he n neve en o level 2. We my hen ee hi gmen wih veice level 3 nd o on. B ince he deh of he ee i onded y N i follow h ome node i eched infiniely ofen y mege fom elow which gnee h hi node i lelled y cceing e of A m infiniely ofen. Th he n i n cceing n of he omon A. Now we n hi ino finie e omon. Fi of ll oeve h he fi comonen of he lel of ny child of n i e of he fi comonen of he lel of n. Fhe he fi comonen of he childen e ll dijoin. Th if we foc j on he fi comonen, we he oo nd we go down, we kee ioning he e ino fine nd fine. Th, he ol nme of node in he ee i onded y 2 n whee Q = n. A ech e, we will dd he mo F n new node. Th, if we hve e of 3n + 1 node (following he ide in Lece 8 o he concion of A d ove) we e gneed h ny node h i doed doe no e in he following e. Wih hi we ge finie e omon, wih Rin cceing condiion wih 3n + 1 i. In he ih i (F i,i i ), F i coni of ll he ee whee n i doe no e nd I i coni of ll he ee whee n i i lelled y n cceing e of A m. Wh i he ize of hi omon? How mny ee of he kind decied ove cn e fomed of ize 2 n fom e of 3n + 1 diinc node. By Cyley heoem he nme of ee ove O(n) node i of he ode of 2 O(nlogn). We hve odeed ee (i.e. diffeen odeing of he childen of ny node yield diffene ee). Fom ech nodeed ee we my genee o n! diffeen lelled ee. Th he ol nme of odeed ee conine o e of he ode of 2 O(nlogn). Given ch ee how mny diffeen wy cn we lel he node wih e of Q ifying he ioning oey? Noe h ech Q e (if 12

ll) long one nie h. Th, if we deemine he il (i.e. he highe nmeed node) fo ech Q, he lelling of he ee i niely deemined. The nme of fncion fom Q o he e of node i n 3n+1. Th he ol nme of Sf ee i O(2 (O(nlogin) ). O eenion of Sf ee i omewh diffeen fom he one ed y Sf in hi e. Thi i ece of o em o ive fom n nlyi of he n-ee. We hll oline he oiginl concion iefly he eginning of he nex lece. Refeence [1] Chiof Löding: Oiml Bond fo Tnfomion of ω-om, Poceeding of he Inenionl Confeence on Fondion of Sofwe Techonology nd Theoeicl Come Science (FSTTCS) 1999, Singe Lece Noe in Come Science 1738, 1999. [2] Mdhvn Mknd: Finie Aom on Infinie Wod, Inenl Reo, SPIC Science Fondion, TCS-96-2, 1996. Aville h://www.cmi.c.in/ mdhvn/e//c-96-2..gz [3] S. Sf: On he comlexiy of ω-om, Poceeding of he 29h FOCS, 1988. [4] Wolfgng Thom: Lngge, om, nd logic In he Hndook of Foml Lngge, volme III, ge 389-455. Singe, New Yok, 1997. 13