A simple algorithm to generate the minimal separators and the maximal cliques of a chordal graph



Similar documents
Recall from Last Time: Disjoint Set ADT

A New Efficient Distributed Load Balancing Algorithm for OTIS-Star Networks

4.1 Interval Scheduling. Chapter 4. Greedy Algorithms. Interval Scheduling. Interval Scheduling: Greedy Algorithms

MPLS FOR MISSION-CRITICAL MICROWAVE NETWORKS BUILDING A HIGHLY RESILIENT MICROWAVE NETWORK WITH MULTI-RING TOPOLOGY

Reading. Minimum Spanning Trees. Outline. A File Sharing Problem. A Kevin Bacon Problem. Spanning Trees. Section 9.6

Cloud and Big Data Summer School, Stockholm, Aug., 2015 Jeffrey D. Ullman

Authenticated Encryption. Jeremy, Paul, Ken, and Mike

Constrained Renewable Resource Allocation in Fuzzy Metagraphs via Min- Slack

Change Your History How Can Soccer Knowledge Improve Your Business Processes?

CPS 220 Theory of Computation REGULAR LANGUAGES. Regular expressions

QUANTITATIVE METHODS CLASSES WEEK SEVEN

Roof Terraces. Structural assemblies

Uses for Binary Trees -- Binary Search Trees

tis, cis cunc - cunc - tis, cis tis, cis cunc - tis, func - def - def - tis, U func - def - func - tis, pa - tri pa - tri pa - tri tu - per - tu -

Matching Execution Histories of Program Versions

Operation Transform Formulae for the Generalized. Half Canonical Sine Transform

Managing Multiple Outsourcing: Service Quality and Volume Issues

ELECTRICAL CAPACITY SYMBOL EQUIPMENT TYPE LOCATION / SERVING MFR MODEL (GALLONS) VOLTS PH AMPS WATTS

11 + Non-verbal Reasoning

CCD CHARGE TRANSFER EFFICIENCY (CTE) DERIVED FROM SIGNAL VARIANCE IN FLAT FIELD IMAGES The CVF method

Online Load Balancing and Correlated Randomness

Form: Parental Consent for Blood Donation

ANALYSIS OF ORDER-UP-TO-LEVEL INVENTORY SYSTEMS WITH COMPOUND POISSON DEMAND

Example A rectangular box without lid is to be made from a square cardboard of sides 18 cm by cutting equal squares from each corner and then folding

Hermes: Dynamic Partitioning for Distributed Social Network Graph Databases

Link-Disjoint Paths for Reliable QoS Routing

Victims Compensation Claim Status of All Pending Claims and Claims Decided Within the Last Three Years

Binary Search Trees. Definition Of Binary Search Tree. Complexity Of Dictionary Operations get(), put() and remove()

PRESENTED TO. Data Leakage Worldwide: The Effectiveness of Corporate Security Policies

Upward Planar Drawings of Series-Parallel Digraphs with Maximum Degree Three

Term Structure of Interest Rates: The Theories

WAVEGUIDES (& CAVITY RESONATORS)

Schedule C. Notice in terms of Rule 5(10) of the Capital Gains Rules, 1993

Finite Dimensional Vector Spaces.

Sun Synchronous Orbits for the Earth Solar Power Satellite System

Econ 371: Answer Key for Problem Set 1 (Chapter 12-13)

Distributed Systems Principles and Paradigms. Chapter 11: Distributed File Systems. Distributed File Systems. Example: NFS Architecture

Fundamentals of Tensor Analysis

(Analytic Formula for the European Normal Black Scholes Formula)

Traffic Flow Analysis (2)

NON-CONSTANT SUM RED-AND-BLACK GAMES WITH BET-DEPENDENT WIN PROBABILITY FUNCTION LAURA PONTIGGIA, University of the Sciences in Philadelphia

Vectors Recap of vectors

NEGLIGENT INFLICTION OF EMOTIONAL DISTRESS IS STILL ALIVE IN TEXAS

Algorithmic Aspects of Access Networks Design in B3G/4G Cellular Networks

Attorney Directory. Prepared for on 12/8/2008. Helping you balance work and life. To use your plan:

1.- L a m e j o r o p c ió n e s c l o na r e l d i s co ( s e e x p li c a r á d es p u é s ).

MAXIMAL CHAINS IN THE TURING DEGREES

Higher. Exponentials and Logarithms 160

Targeted Cancer Gene Analysis For Research Use Only - Not Intended for Clinical Use

The example is taken from Sect. 1.2 of Vol. 1 of the CPN book.

Yuriy Alyeksyeyenkov 1

Inductive Proximity Sensors For Reliable Precision Feedback

Section 2.2 Future Value of an Annuity

VOL. 25, NÚM. 54, EDICIÓN JUNIO 2007 PP

Applying the actuarial control cycle in private health insurance

Business rules FATCA V. 02/11/2015

December Homework- Week 1

MODULE 3. 0, y = 0 for all y

SecurView Antivirus Software Installation


News English.com Ready-to-Use English Lessons by Sean Banville Level 6 Students sell bodies and gamble to survive

Quality and Pricing for Outsourcing Service: Optimal Contract Design

No 28 Xianning West Road, Xi an No 70 Yuhua East Road, Shijiazhuang.

Section 3: Logistic Regression

SAN JOSE UNIFIED RETURNING VOLUNTEER DRIVER PACKET

Graph Theoretical Analysis and Design of Multistage Interconnection Networks

Mathematics. Vectors. hsn.uk.net. Higher. Contents. Vectors 128 HSN23100

Binary Search Trees. Definition Of Binary Search Tree. The Operation ascend() Example Binary Search Tree

Important result on the first passage time and its integral functional for a certain diffusion process

A Place to Choose Quality, Affordable Health Insurance

Earthquake Hazard Zones: The relative risk of damage to Canadian buildings

A Formal Model for Data Flow Diagram Rules

Sharp bounds for Sándor mean in terms of arithmetic, geometric and harmonic means

OHIM SEARCH TOOLS: TMVIEW, DSVIEW AND TMCLASS. Making trade mark and design information readily available for users

- ASSEMBLY AND INSTALLATION -

1.5 / 1 -- Communication Networks II (Görg) Transforms

DYNAMIC PROGRAMMING APPROACH TO TESTING RESOURCE ALLOCATION PROBLEM FOR MODULAR SOFTWARE

Menu Structure. Section 5. Introduction. General Functions Menu

CHAPTER FIVE. Solutions for Section 5.1. Skill Refresher. Exercises

With Rejoicing Hearts/ Con Amor Jovial. A Fm7 B sus 4 B Cm Cm7/B


Elements from Another Universe: Understanding the Beauty of the Periodic Table

Back left Back right Front left Front right. Blue Shield of California. Subscriber JOHN DOE. a b c d

Factoring Polynomials: Factoring by Grouping

C e r t ifie d Se c u r e W e b

SE3BB4: Software Design III Concurrent System Design. Sample Solutions to Assignment 1

The Mathematics of Sudoku

5.4 Exponential Functions: Differentiation and Integration TOOTLIFTST:

Evobike 2014 Árak szerzõdött partnereink számára Mounty

Criminal Offenses - On campus

Modern Portfolio Theory (MPT) Statistics

Connector and Cable Diagrams (Pinout Charts)

Engineering. Installation Manager. Installations Manager (Gas Qualified) Whole House Assessors. Whole House Assessors (Gas Qualified)

Path-dependent options

Director s Statement

Transcription:

A smpl lgortm to gnrt t mnml sprtors nd t mxml lqus o ordl grp Ann Brry 1 Romn Pogorlnk 1 Rsr Rport LMOS/RR-10-04 Fbrury 11, 20 1 LMOS UMR CNRS 6158, Ensmbl Sntqu ds Cézux, F-63 173 Aubèr, Frn, brry@sm.r

Abstrt W prsnt smpl und lgortm pross w uss tr LxBFS or MCS on ordl grp to gnrt t mnml sprtors nd t mxml lqus n lnr tm n sngl pss. Kywords: MCS, LxBFS, mnml sprtor, mxml lqu, ordl grp. 1

1 ntroduton Gnrtng t mnml sprtors nd t mxml lqus o ordl grp s smpl wt t lp o sp sr lgortms. Cordl grps (w r t grps wt no ordlss yl on our or mor vrts), wr rtrzd by Fulkrson nd Gross [7] s t grps or w on n rptdly nd smpll vrtx (.. vrtx wos ngborood s lqu) nd rmov t rom t grp, untl no vrtx s lt; ts pross, lld smpll lmnton, dns n ordrng α on t vrts lld prt lmnton ordrng (po). (α() dnots t vrtx brng numbr, nd α 1 (x) t numbr o vrtx x). At stp o t lmnton pross, nw trnstory grp s dnd. Ros [] sowd tt or ny gvn po α o ordl grp, ny mnml sprtor o t grp s dnd n t ours o t smpll lmnton pross s t trnstory ngborood o som vrtx. Morovr, t s sy to s tt n ny mxml lqu K, t vrtx y o smllst numbr by α dns K s ts trnstory losd ngborood. n t rst, w wll ll su vrts gnrtors, nd our gol wll b to dtt ts gnrtors ntly. W wll us grp sr lgortms LxBFS [10] nd MCS [11], bot tlord to gnrt po o ordl grp. T d bnd ts s qut smpl. Bot lgortms numbr t vrts rom n to 1 (n s t numbr o vrts o t grp). E vrtx brs lbl, w my b modd s t lgortm prods. At stp, nw vrtx s osn to b numbrd. t lbl o ts nw vrtx x s not lrgr tt t lbl o t prvous vrtx x +1, tn w know tt x gnrts mnml sprtor o t grp. T vrts o ts mnml sprtor r t lrdy numbrd ngbors o x. W lso know tt t prvously numbrd vrtx, x +1, gnrts mxml lqu: x +1, togtr wt ts lrdy numbrd ngbors (xludng x ), dn mxml lqu o t grp. Tus t lbls o lgortms LxBFS nd MCS nbl t usr to dtt ts gnrtors s soon s ty r numbrd. T mnml sprtors nd mxml lqus n b ound on-ln, wtout rqurng prlmnry pss o t lgortm to numbr ll t vrts. T orrspondng rsults v bn provd or MCS n two sprt pprs. Blr nd Pyton [5], wl studyng ow MCS dns lqu tr o ordl grp, sowd ow to gnrt t mxml lqus (w r t nods o t lqu tr), usng t MCS lbls. Kumr nd Mdvn [8] sowd ow to gnrt t mnml sprtors o ordl grp, gvn n MCS ordrng. 2

Howvr, ltoug mnml sprtor nd mxml lqu gnrtor r tully dttd t t sm stp o t lgortm, ts rsults v not bn und. Morovr, [8] us po s nput, but t mnml sprtors s wll s t mxml lqus n b omputd durng t xuton, w my b n mportnt tur wn ndlng vry lrg grps, sn globl pr-numbrng o t vrts s not nssry or vn usul. LxBFS, s w wll sow, xbts t sm proprty s MCS rgrdng ts gnrtors. Our m s to prsnt smpl und sngl-pss lgortm w gnrts t mnml sprtors nd t mxml lqus o ordl grp. T ppr s orgnzd s ollows: n Ston 2, w gv som prlmnry rsults nd dntons. n Ston 3, w prsnt our mn torm, w dsrbs ow t gnrtors r dttd, nd dsuss ts proo or LxBFS. Ston 4 prsnts our lgortm nd provds n xmpl. 2 Prlmnrs n t rst, w wll onsdr ll grps to b onntd (or dsonntd grp, t prosss w dsrb n b ppld ndpndntly to onntd omponnt). A grp s dnotd G = (V, E), V = n, E = m. Vrts x nd y r djnt xy E. For X V, G(X) dnots t subgrp ndud by X. T ngborood N G (x) o vrtx x n grp G s N G (x) = {y x xy E} (subsrpts wll b omttd wn t s lr w grp w work on). T losd ngborood s N G [x] = {x N(x)}. T ngborood o vrtx st X V s N G (X) = x X N G (x) X. A subst X o vrts s lld modul t vrts o X sr t sm xtrnl ngborood: x X, N(x) X = N(X). A lqu s st o prws djnt vrts. A subst S o vrts o onntd grp G s lld sprtor G(V S) s not onntd. A sprtor S s lld n xy-sprtor x nd y l n drnt onntd omponnts o G(V S), mnml xy-sprtor S s n xy-sprtor nd no propr subst o S s n xy-sprtor. A sprtor S s mnml sprtor, tr s som pr {x, y} su tt S s mnml xy-sprtor. Altrntly, S s n mnml sprtor nd only G(V S) s t lst 2 onntd omponnts C 1 nd C 2 su tt N(C 1 ) = N(C 2 ) = S; su omponnts r lld ull omponnts o S n G, nd S s tn mnml xy-sprtor or ny {x, y} wt x C 1 nd y C 2. LxBFS [10] nd MCS [11] r bot lnr-tm sr lgortms w numbr t vrts rom n to 1, trby dnng po. E vrtx x brs lbl w orrsponds to t lst o numbrs o t ngbors o x wt 3

gr numbr tn tt o x (LxBFS) or to t rdnlty o ts lst (MCS). Bot lgortms r rlld blow. Algortm LxBFS (Lxogrp Brdt-Frst Sr)[10] nput : A grp G = (V, E). output: An ordrng α o V. ntlz ll lbls s t mpty strng; or n to 1 do Pk n unnumbrd vrtx v wos lbl s mxml undr lxogrp ordr; α() v ; or unnumbrd vrtx w djnt to v do ppnd to lbl(w); Algortm MCS (Mxml Crdnlty Sr)[11] nput : A grp G = (V, E). output: An ordrng α o V. ntlz ll lbls s 0; or n to 1 do Pk n unnumbrd vrtx v wt mxmum lbl; α() v ; or unnumbrd vrtx w djnt to v do lbl(w) lbl(w) + 1; 3 Mn Torm W wll now prsnt our mn torm, rom w our lgortm s drvd. W wll nd t ollowng dntons: Dnton 3.1 Gvn ordl grp G nd po α o G, or ny vrtx x, Mdj(x) s t st o ngbors o x wt numbr gr tn tt o x: Mdj(x) = {y N(x) α 1 (y) > α 1 (x)}. Dnton 3.2 Gvn ordl grp G nd po α o G, w wll ll: 4

mnml sprtor gnrtor ny vrtx x wt numbr by α, su tt Mdj(x ) s mnml sprtor o G nd lbl(x ) lbl(x +1 ), wr x +1 = α( + 1). mxml lqu gnrtor vrtx y su tt Mdj(y) {y} s mxml lqu o G. Our mn rsult s t ollowng: Torm 3.3 Lt α b po dnd by tr LxBFS or MCS, lt x b t vrtx o numbr, lt x +1 b t vrtx o numbr + 1. ) x s mnml sprtor gnrtor nd only lbl(x ) lbl(x +1 ). b) x +1 s mxml lqu gnrtor nd only lbl(x ) lbl(x +1 ) or = 1. W wll now dsuss t proo o Torm 3.3. W wll rst prsnt moplx lmnton, w s pross tt xplns ow bot MCS nd LxBFS sn t mnml sprtors nd t mxml lqus o ordl grp, tn prov Torm 3.3 or LxBFS. 3.1 Moplx lmnton W wll us t noton o moplx, ntrodud by [1]: Dnton 3.4 [1] A moplx s lqu X su tt X s modul nd N(X) s mnml sprtor. W xtnd ts dnton to lqu wos ngborood s mpty. A smpll moplx s moplx wos vrts r ll smpll. [1] provd: Proprty 3.5 [1] Any ordl grp w s not lqu s t lst two non-djnt smpll moplxs. From ts, ty drvd vrnt o t rtrzton o Fulkrson nd Gross or ordl grps by smpll lmnton o vrts: Crtrzton 3.6 [1] A grp s ordl nd only on n rptdly dlt smpll moplx untl t grp s lqu (w w wll ll t trmnl moplx ). W ll ts pross moplx lmnton. 5

Not tt moplx lmnton on ordl grp s spl s o smpll lmnton, sn t stp st o smpll vrts s lmntd. Not lso tt or onntd grp G, t trnstory lmnton grp obtnd t t nd o stp rmns onntd. Moplx lmnton dns n ordrd prtton (X 1, X 2,..., X k ) o t vrts o t grp nto t sussv moplxs w r dnd n t sussv trnstory lmnton grps. W wll ll ts prtton moplx ordrng. Our bs rsult s t ollowng: Torm 3.7 1 Lt G b ordl grp, lt (X 1, X 2,..., X k ) b moplx ordrng o G. At stp < k o t lmnton pross ndng moplx X n trnstory grp G, N G (X ) s mnml sprtor o G. X N G (X ) s mxml lqu o G. T trmnl moplx X k s mxml lqu. Tr r no otr mnml sprtors or mxml lqus n G. To prov ts, w wll rst rll t rsult rom Ros []: Proprty 3.8 [] Lt G b ordl grp, lt α b po o G, lt S b mnml sprtor o G. Tn tr s som vrtx x su tt Mdj(x) = S. From Proprty 3.8, t s sy to ddu t ollowng wll-known proprty: Proprty 3. [] Lt G b ordl grp, lt S b mnml sprtor o G. Tn n vry ull omponnt C o S, tr s som vrtx x su tt S N(x) (su vrtx s lld onlun pont o C). Lt us now prov Torm 3.7. Proo: (o Torm 3.7) Lt us rst prov tt N G (X ) s mnml sprtor o G. n trnstory grp G, N G (X ) s by dnton mnml sprtor S o G, wt t lst two ull omponnts C 1 nd C 2, ontnng onlun pont, w w wll ll x 1 nd x 2. Suppos S s not mnml sprtor o G. Tn tr must b ordlss pt rom x 1 to x 2 n G w ontns no vrtx o S. Lt y b t rst vrtx o ts pt to b lmntd, t 1 T rsults provd n ts subston wr brly prsntd n [2]. 6

stp j < ; y must b smpll n G j, but ts s mpossbl, sn y s two non-djnt ngbors on t pt. Tus vry N G (X ) ( < k) s mnml sprtor o G, nd by Proprty 3.8, ll t mnml sprtors o G v tus bn nountrd. Lt us now prov tt X N G (X ) s mxml lqu o G, by nduton on. Clrly, X 1 N G (X 1 ) s mxml lqu o G. Lt us xmn X N G (X ): t s mxml lqu o G nd tus lqu o G. Suppos t s not mxml lqu o G, nd lt y b vrtx blongng to lrgr lqu o G ontnng X N G (X ). Sn N G (X ) s mnml sprtor o G, y must blong to t sm ull omponnt o N G (X ) s t vrts o X. But tn y blongs to t sm mxml lqu modul s X, ontrdton. T trmnl moplx X k s by dnton lqu; suppos t s not mxml lqu: tn t n b ugmntd wt som vrtx y, w blongs to som moplx X. By Torm 3.7, S = N G (X ) s mnml sprtor o G. Lt C 1,..., C t b t ull omponnts o S n G, nd lt z b t onlun pont blongng to t moplx o gst numbr X j. n G j, S nnot b mnml sprtor, sn ll t otr ull omponnts o S v bn lmntd rom G j. Tror, X j N Gj (X j ) must b t trmnl moplx, ontrdton. Not tt or gvn ordl grp G, tr my b mny drnt moplx ordrngs, but tr s lwys t sm numbr o moplxs, sn ts s t numbr o mxml lqus o t grp. Wt ny moplx ordrng (X 1, X 2,..., X k ), w n ssot po α by prossng t moplxs rom 1 to n nd gvng onsutv numbrs to t vrts o gvn moplx. Usng α, w n dn t mnml sprtors nd mxml lqus s vrtx ngboroods: Proprty 3.10 Lt G b ordl grp, lt (X 1, X 2,..., X k ) b moplx ordrng o G, lt α b po ssotd wt ts moplx ordrng. Tn n t ours o moplx lmnton, t stp, prossng moplx X, T vrtx x o moplx X wos numbr s t smllst by α dns mxml lqu {x} Mdj(x) o G. T vrtx y o moplx X wos numbr s t gst by α dns mnml sprtor Mdj(y) o G. 7

3.2 Gnrtors o LxBFS LxBFS dns moplx ordrng nd n ssotd po: [1] sowd tt LxBFS lwys numbrs s 1 vrtx blongng to moplx (w w wll ll X 1 ). Ty lso provd tt t vrts o X 1 rv onsutv numbrs by LxBFS. Ts proprts r tru t stp o LxBFS n t trnstory lmnton grp. Tror, LxBFS dns moplx lmnton (X 1,..., X k ), by numbrng onsutvly t vrts o X 1, tn numbrng onsutvly t vrts o X 2, nd so ort: Torm 3.11 n ordl grp, LxBFS dns moplx ordrng. Not tt t s sy to ddu rom [5] tt MCS lso dns moplx ordrng. Exmpl 3.12 Fgur 1 sows t numbrs nd lbls o LxBFS xuton on ordl grp. moplx mnml sprtor mxml lqu X 1 = {g} {b, } {b,, g} X 2 = {} {b,, d} {, b,, d} X 3 = {d, } {b, } {b,, d, } X 4 = {} {b, } {b,, } X 5 = {b,, } - {b,, } 2 [6,5,4] [] 8 [8,7] 6 b [7,6] 5 [6,5] 4 d 7 [8] g 1 [6,5] [6,5,4] 3 Fgur 1: Numbrs nd lbls o LxBFS xuton on ordl grp. Sn t vrts o ny trnstory moplx X r numbrd onsutvly, t lbls wll nrs w.r.t. lxogrp ordr untl t ntr moplx s bn numbrd, nd tn wll stop nrsng. Mor prsly, wn 8

runnng LxBFS (numbrng t vrts rom n to 1), s long s t lbls nrs, w r dnng moplx, X. Wn t lbls stop nrsng (wn numbrng vrtx y), tn w v strtd nw moplx X 1 w wll ontn y. Usng Proprty 3.10, w n ddu tt LxBFS gnrts t mnml sprtors nd t mxml lqus ordng to Torm 3.3. Tus t s sy, usng LxBFS ordrng, to dn wt sngl pss t mnml sprtors o G t mxml lqus o G t orrspondng moplx ordrng 4 Algortm W n now drv rom Torm 3.3 gnrlzd lgortm to gnrt t mnml sprtors nd t mxml lqus o ordl grp. n t lgortm blow, t s onsdrd tt tr LxBFS or MCS s usd. Us to nrmnt t lbl o y s trnsltd s dd to lbl o y or LxBFS nd s dd 1 to t lbl o y or MCS. T lbls r ll onsdrd ntlzd t t bgnnng, s or LxBFS nd s 0 or MCS. G NUM G NUM +{x } s sortnd or V NUM V NUM +{x }; G NUM G(V NUM ). n t sm son, G ELM G ELM {x } s sortnd or V ELM V ELM {x } ; G ELM G(V ELM ) ; Symbol + dnots dsjont unon. Not tt or LxBFS, x s mnml sprtor gnrtor, tn t lbl o x gvs t sprtor drtly., or nstn, lbl(x ) = (, 7, 6), tn t mnml sprtor w x gnrts wll b {x, x 7, x 6 }.

Algortm Mnsps-Mxlqus nput : A ordl grp G = (V, E). output: St S o mnml sprtors o G ; St S o mnml sprtor gnrtors o G ; St K o mxml lqus o G ; St K o mxml lqu gnrtors o G ; nt: G NUM G( ); G ELM G ; S ; S ; K ; K ; or =n downto 1 do Coos vrtx x o G ELM o mxmum lbl ; G NUM G NUM + {x } ; n nd lbl(x ) λ tn //x s mn. sp. gnrtor nd x +1 s mx. lqu gnrtor S S + {x } ; S S {N GNUM (x )} ; K K + {x +1 } ; K K + {(N GNUM (x ) {x }}; λ lbl(x ) ; or y N GELM (x ) do Us to nrmnt t lbl o y ; G ELM G ELM {x } ; K K + {x 1 } ; K K + {N G (x 1 )} ; T omplxty o t bov lgortm s t sm s or LxBFS or MCS, w s lnr (O(n + m)) tm. Exmpl 4.1 Fgur 4 blow gvs stp-by-stp xmpl usng MCS. St o mnml sprtors: S = {{}, {g, }, {d}}. St o mnml sprtor gnrtors: S = {d,, }. St o mxml lqus: K = {{, b, }, {, d, g, }, {g,, }, {d,, }}. St o mxml lqu gnrtors: K = {,,, }. Wn prossng vrtx x 6, x 6 = d s dnd s mnml sprtor gnrtor or {} nd tus x 7 = s mxml lqu gnrtor or {, b, }. Wn prossng vrtx x 3, x 3 = s dnd s mnml sprtor gnrtor or {g, }, so x 4 = s mxml lqu gnrtor, or {, d, g, }. Wn prossng vrtx x 2, x 2 = s dnd s mnml sprtor gnrtor or {d}, so x 3 = s mxml lqu gnrtor or {g,, }. 10

At t nd, x 1 = s dnd mxml lqu gnrtor, or {d,, }. b d 8 b d 7 8 b d g g g Prossng Prossng 8 Prossng 7 7 8 b 00 11 d 00 116 00 11 7 8 b d 6 7 8 b d 6 g 5 g 5 g 4 Prossng 6 Prossng 5 Prossng 4 7 8 b d 6 5 g 4 7 8 2 b d 00 11 00 11 6 5 g 4 7 8 b 2 1 00 11 d 00 11 00 11 6 5 g 4 3 3 Prossng 3 Prossng 2 Prossng 1 3 Fgur 2: A stp by stp xuton usng MCS. T dsd vrtx numbrs rprsnt t mnml sprtor gnrtors, nd t rld vrtx sts r t mnml sprtors. 4.1 Multplty o t gnrtd mnml sprtors A mnml sprtor my b gnrtd svrl tms, dpndng on t numbr o ull omponnts t dns: Proprty 4.2 Lt α b po dnd by LxBFS or MCS, lt S b mnml sprtor, lt k b t numbr o ull omponnts o S. Tn S s k 1 gnrtors by α. 11

Proo: n t ours o smpll lmnton pross on smpll vrts, bor ny vrtx s o mnml sprtor S n b lmntd, ll t ull omponnts o S (xpt on) must v bn lmntd, ls s nnot b smpll, s s ss t onlun ponts o rmnng ull omponnt. T lst vrtx lmntd rom ts ull omponnts gnrts S. Wn ll ull omponnts (xpt on) v bn lmntd, S s no longr mnml sprtor. Tror S s k ull omponnts, t s gnrtd xtly k 1 tms. Not tt [8] gvs n O(n + mlogn) pross to nd t multplty o mnml sprtor (wt MCS). Howvr, ty lso prov tt wt LxBFS, gvn t mnml sprtor gnrtor o lowst numbr w gnrts S, wtn t subst o mnml sprtor gnrtors w dn mnml sprtor o sz S, ll t gnrtors o S r onsutv. n t ours o n xuton o LxBFS, on n stor t lbls o t gnrtors n drnt lsts ordng to t sz o t mnml sprtor ty gnrt. Tn on n run troug lst, tstng or onsutv ourrns o gvn mnml sprtor. Ts n b don n lnr tm, s by Torm 3.7, t sum o t szs o t mnml sprtors s lss tn m. 5 Conluson n ts ppr, w prsnt smpl nd optml pross w gnrts t mnml sprtors nd mxml lqus o ordl grp n sngl pss o tr LxBFS or MCS, wtout rqurng t prlmnry omputton o po. Toug bot LxBFS nd MCS yld n optml lnr-tm pross or ts problm, t s mportnt to not tt ty dn drnt st o pos o ordl grp [4], nd xbt drnt lol bvors [3]. t my b mportnt to us on or t otr, dpndng on t ntndd pplton. W wll nd ts ppr by notng tt two rntly ntrodud sr lgortms, LxDFS nd MNS [6] l to bv n t sm son. A ountr-xmpl or MNS s gvn blow (T MNS lbls r t sts o gr-numbrd ngbors, omprd by st nluson): wn vrtx 2 s numbrd, ts lbl s not grtr tn tt o vrtx 3; owvr, Mdj(3) = {3, 4} s not mxml lqu.. 12

{4} 3 {5} 4 5 2 {5} 1 {4,3} Fgur 3: MNS dos not gnrt t mxml lqus. Rrns [1] A. Brry, J.-P. Bordt Sprblty gnrlzs Dr s torm. Dsrt Appld Mtmts, 84(1-3):43 53, 18. [2] A. Brry, J.-P. Bordt Moplx lmnton ordrngs. Eltron Nots n Dsrt Mtmts, 8:6, 20. [3] A. Brry, J. R. S. Blr, J.-P. Bordt, nd G. Smont. Grp xtrmts dnd by sr lgortms. Rsr rport, LMOS, RR-07-05, 2007, to ppr n Algortms. [4] A. Brry, R. Krugr, G. Smont Gnvèv Smont. Mxml lbl sr lgortms to omput prt nd mnml lmnton ordrngs. SAM J. Dsrt Mt., 23(1):428 446, 200. [5] J. R. S. Blr nd B. W. Pyton. An ntroduton to ordl grps nd lqu trs. Grp Tory nd Sprs Mtrx Computton. [6] D. G. Cornl, R. Krugr A und vw o grp srng. SAM J. Dsrt Mt., 22(4):125 1276, 2008. [7] D. R. Fulkrson nd O. A. Gross. ndn mtrs nd ntrvl grps. P Journl o Mtmts, 28:565 570, 16. [8] P. Srnvs Kumr nd C. E. Vn Mdvn. Mnml vrtx sprtors o ordl grps. Dsrt Appld Mtmts, 8(1-3):155 168, 18. [] D. J. Ros. Trngultd grps nd t lmnton pross. Journl o Mtmtl Anlyss nd Appltons,. [10] D. J. Ros, R. E. Trjn, nd G. S. Lukr. Algortm spts o vrtx lmnton on grps. SAM J. Comput., 5(2):266 283, 176. 13

[11] R. E. Trjn nd M. Ynnkks. Smpl lnr-tm lgortms to tst ordlty o grps, tst ylty o yprgrps, nd sltvly rdu yl yprgrps. SAM J. Comput., 13(3):566 57, 184. 14