• Aucun résultat trouvé

Turing machines application to the rate of decrease in logical depth for general Logical depth for reversible Turing machines with an Theoretical Computer Science

N/A
N/A
Protected

Academic year: 2022

Partager "Turing machines application to the rate of decrease in logical depth for general Logical depth for reversible Turing machines with an Theoretical Computer Science"

Copied!
3
0
0

Texte intégral

(1)

Theoretical Computer Science 778 (2019) 78–80

Contents lists available atScienceDirect

Theoretical Computer Science

www.elsevier.com/locate/tcs

Note

Logical depth for reversible Turing machines with an

application to the rate of decrease in logical depth for general Turing machines

Paul M.B. Vitányi

a

,

b

,∗

aCWI,theNetherlands

bUniversityofAmsterdam,theNetherlands

a rt i c l e i n f o a b s t ra c t

Articlehistory:

Received16January2019 Accepted17January2019 Availableonline25January2019 CommunicatedbyP.G.Spirakis

Keywords:

Logicaldepth Kolmogorovcomplexity Compression

The logical depth of a reversible Turing machine equals the shortest running time of a shortest program for it. This is applied to show that the result in [1] is valid notwithstandingtheerrornotedinCorrigendum [7].

©2019ElsevierB.V.Allrightsreserved.

1. Introduction

Abookonnumbertheoryisdifficult,or‘deep.’Thebooklistsanumberofdifficulttheoremsofnumbertheory.However, it hasvery low Kolmogorov complexity,sinceall theorems arederivable fromtheinitial few definitions.Ourestimate of thedifficulty,or‘depth,’ofthebookisbasedonthefactthat ittakesalongtime toreproducethebookfrompartofthe informationinit.Theexistenceofadeepbookisitselfevidenceofsomelongevolutionprecedingit.

Thelogicaldepthofa(finite)stringisrelatedtocomplexitywithboundedresourcesandmeasuresthetradeoffbetween program sizesandrunningtimes.Computinga stringxfromoneofitsshortestprogramsmaytakea verylongtime,but computingthesamestringfroma simple“print(x)” programoflengthabout|x|bits (thelengthofx) takesverylittle time.

Logical depth asdefinedin [4] for a string comes in two versions: one based on thecompressibility of programs of prefixTuringmachinesandtheotherusingtheratiobetweenalgorithmicprobabilitieswithandwithouttimelimits.Since both areapproximatelythesame([5,Theorem7.7.1] basedon[4,Lemma3])itisnorestrictiontousethecompressibility version.

Theusednotionsofcomputability,resource-boundedcomputation time,self-delimitingstrings,big-Onotation,andKol- mogorovcomplexityarewell-knownandtheproperties,notations,aretreatedin[5].

*

Correspondenceto:CWI,SciencePark123,1098XGAmsterdam,theNetherlands.

E-mailaddress:Paul.Vitanyi@cwi.nl.

https://doi.org/10.1016/j.tcs.2019.01.031

0304-3975/©2019ElsevierB.V.Allrightsreserved.

(2)

P.M.B. Vitányi / Theoretical Computer Science 778 (2019) 78–80 79

2. Preliminaries

AllTuringmachinesinthispaperareprefixTuringmachines.AprefixTuringmachineisaTuringmachinewithaone-way read-onlyprogramtape,anauxiliarytape,oneormoreworktapesandanoutputtape.Alltapesarelinearone-wayinfinite anddividedintocellscapableofcontaining onesymbolout ofa finiteset.Initially theprogramtapeisinscribedwithan infinitesequenceof0’sand1’sandtheheadisscanningtheleftmostcell.Whenthecomputationterminatesthesequence ofbitsscanned ontheinputtape istheprogram.Foreveryfixed finitecontentsoftheauxiliary tapethesetofprograms forsuch a machine isa prefix code(no program isa proper prefix of another program). Let T0

,

T1

, . . .

be the standard enumeration ofprefix Turingmachines. A universalprefixTuringmachine simulates every prefix Turing machine givenits indexnumber.Wealsorequireittobeoptimalwhichmeansthatthesimulationprogramisasshortaspossible.Wechoose areferenceoptimaluniversalprefixTuringmachineandcallitU.

Theprefix Kolmogorovcomplexity isbasedontheprefixTuringmachinesimilarto the(plain)Kolmogorov complexity based on the(plain) Turing machine.Let x

,

y be finite binary strings. The prefixKolmogorovcomplexity K

(

x|y

)

of x with auxiliary y isdefinedby

K

(

x

|

y

) =

min

p

{|

p

| :

U

(

p

,

y

) =

x

}.

Ifxis abinary stringoflengthnthen K

(

x|y

)

n+O

(

logn

)

.Restrictingthe computationtime resourceisindicated bya superscriptgivingthe allowed numberofsteps, usually denotedby d.The notation Ud

(

p

,

y

)

=x meansthat U

(

p

,

y

)

=x withindsteps.Iftheauxiliarystringy istheemptystring ,thenweusuallydropit.Similarly,wewriteU

(

p

)

forU

(

p

, )

. Thestringx isashortestprogramforxifU

(

x

)

=xandK

(

x

)

= |x|.Astringxisb-incompressibleif|x|≥ |x|−b.

3. ReversibleTuringmachines

ATuringmachinebehavesaccordingtoafinitelistofrules.Theserulesdetermine,fromthecurrentstateofthefinite controlandthesymbolcontainedinthecellunderscan,theoperationtobeperformednextandthestatetoenteratthe endofthenextoperationexecution.

Thedevice is(forward) deterministic.Notevery possiblecombinationofthe firsttwoelements hastobe intheset;in thiswaywe permitthedevice toperformnooperation. Inthiscasewesaythat thedevicehalts.Hence,wecandefine a Turingmachinebyatransitionfunction.

Definition1.A reversible Turing machine[3,2] is a Turing machinethat is forwarddeterministic (any Turing machine as definedis)butalsobackwarddeterministic,thatis,thetransitionfunctionhasasingle-valuedinverse.Thedetailsofthefor- maldefinitionareintricate[3,2] andneednotconcernushere.Thisdefinitionextendsintheobviousmannertomultitape Turingmachines.

In[3] forevery 1-tapeordinaryTuringmachine T a3-tapereversibleTuringmachineTrevisconstructedthatemulates T inlineartimesuchthatwithinput ptheoutputisTrev

(

p

)

=

(

p

,

T

(

p

))

.ThereversibleTuringmachinethatemulatesU is calledUrev.

Definition2.Letxbeastringandbanonnegativeinteger.Thelogicaldepthofx atsignificancelevel bis depthb

(

x

) =

min

d

:

p

∈ {

0

,

1

}

Ud

(

p

) =

x

∧ |

p

| ≤

K

(

p

) +

b

,

theleastnumberofstepstocompute xbyab-incompressibleprogram.

Theorem1.Thelogicaldepthofastringx atsignificancelevelbN forreversibleTuringmachinesisequalto depthbrev

(

x

) =

min

d

:

p

∈ {

0

,

1

}

Urevd

(

p

) = (

p

,

x

) ∧ |

p

| ≤

K

(

x

) +

b

,

theleastnumberofstepstocompute

(

p

,

x

)

byUrevfromaprogramoflengthatmostK

(

x

)

+b.

Proof. SinceareversibleTuringmachineisbackwardsdeterministic,andan incompressibleprogram cannotbecomputed froma shorter program,the length ofan incompressibleprogram withb=0 for xcan only be the lengthof a shortest programforx.Thelogicaldepthatsignificanceb isthentheleastnumberofstepstocomputexbyUrevfromaprogramp oflengthatmostK

(

x

)

+b. 2

(3)

80 P.M.B. Vitányi / Theoretical Computer Science 778 (2019) 78–80

4. Therateofdecreaseoflogicaldepth

InCorrigendum [7] thefollowingisobserved:In[1,Section 4] itisassumedthat,forallx∈ {0

,

1},thestringx isthe only incompressiblestringsuch thatU

(

x

)

=x.Thatis,logicaldepth resemblingtheone accordingto Theorem1isused.

However, thisassumptioniswrongforgeneralTuringmachinesinthat formanyxtheremaybeanincompressiblestring pwith|x|≥ |p|

>

|x|suchthatU

(

p

)

=x.ThecomputationofU

(

p

)

=xmaybefasterthanthatofU

(

x

)

=x.Forexample, the functionfromx∈ {0

,

1} totheleastnumberofsteps inacomputation U

(

p

)

=xforan incompressiblestring pmay be computable. The argument in the paper is, however, correct forthe set ofreversible Turing machines. These Turing machinesareasubsetofthesetofallTuringmachines[3,2] andemulatetheminlineartime.Thisimpliesthecorrectness of[1,Theorem2] asweshallshow.

Lemma1.Let

ψ

bedefinedby

ψ (

n

) =

max

|x|=nmin

d

{

d

:

Udrev

(

x

) = (

x

,

x

)}.

Then

ψ

isnotcomputableandgrowsfasterthananycomputablefunction.

Proof. Ifa function

ψ

asinthe lemma were computable,then foran x oflength n we couldrun Urev emulating U [3]

forwardfor

ψ(

n

)

stepsonallprogramsoflengthn+O

(

logn

)

.Amongthoseprogramsthathaltwithin

ψ(

n

)

steps,wecould selectthe programs p whichoutput

(

p

,

x

)

.Subsequently,we could selectfromthat setaprogram p ofminimumlength, say x.Sucha programx haslength K

(

x

)

sinceUrev isemulating U.Thiswouldimplythat K wouldbecomputable.But thefunction K isincomputable[6,5]:contradiction.Therefore

ψ

cannotbecomputable.Sincethisholdsforeveryfunction majoring

ψ

,thefunction

ψ

mustgrowfasterthananycomputablefunction. 2

Corollary1.ThesetofreversibleTuringmachinesisasubsetofthesetofallTuringmachines.TheemulationofU

(

p

)

byUrev

(

p

)

is lineartimeforallbinaryinputsp by[3].Therefore,replacinginthelemmaUrevbyU changes

ψ(

n

)

to

φ (

n

)

=

(ψ(

n

))

.Hencethe lemmaholdswith

ψ

replacedby

φ

andUrevbyU .Thisgivesus[1,Lemma1] andtherefore[1,Theorem2](theBusyBeaverupper boundisprovedasitisin[1]):

Theorem2.Thefunction f

(

n

) =

max

|x|=n,0bn

{

depthb

(

x

)

depthb+1

(

x

)}

growsfasterthananycomputablefunctionbutnotasfastastheBusyBeaverfunction.

(Thestatementof[1,Theorem2] containsatypo“x:whichshouldbedeletedandthesuperscript“

(

2

)

”is(canbe)omittedhere.) References

[1]L.F.Antunes,A.Souto,P.M.B.Vitányi,Ontherateofdecreaseinlogicaldepth,Theoret.Comput.Sci.702(2017)60–64.

[2]H.B.Axelsen,R.Glück,AsimpleandefficientuniversalreversibleTuringmachine,in:Proc.5thInt.Conf.LanguageandAutomataTheoryandApplica- tions,in:LectureNotesinComputerScience,vol. 6638,Springer,2011,pp. 117–128.

[3]C.H.Bennett,Logicalreversibilityofcomputation,IBMJ.Res.Develop.17 (6)(1973)525–532.

[4]C.H.Bennett,Logicaldepthandphysicalcomplexity,in:R.Herken(Ed.),TheUniversalTuringMachineaHalf-CenturySurvey,OxfordUniversityPress, 1988,pp. 227–257.

[5]M.Li,P.M.B.Vitányi,AnIntroductiontoKolmogorovComplexityandItsApplications,Springer,NewYork,2008.

[6]A.N.Kolmogorov,Threeapproachestothequantitativedefinitionofinformation,Probl.Inf.Transm.1 (1)(1965)1–7.

[7]P.M.B.Vitányi,Corrigendumto“Ontherateofdecreaseinlogicaldepth”byL.F.Antunes,A.Souto,andP.M.B.Vitányi[Theoret.Comput.Sci.702(2017) 60–64],Theoret.Comput.Sci.770(2019)101.

Références

Documents relatifs

On the other hand, for Turing machines that halt on the conjectured final loop of the 3x + 1 function, more researches are still to be

From this data, it is possible to deduce the tape symbol of the visited cells and the relative position of the head at each time step.. In order to recognize S T , we can register

Unite´ de recherche INRIA Rocquencourt, Domaine de Voluceau, Rocquencourt, BP 105, 78153 LE CHESNAY Cedex Unite´ de recherche INRIA Sophia-Antipolis, 2004 route des Lucioles, BP

Moreover, there is an infinite sequence of strings such that the minimal computation times of associated shortest programs rises incomputably fast but not so fast as the Busy

The computation times of shortest programs can similarly rise faster than any computable function but not as fast as the Busy Beaver

We formalize Turing’s description as follows: A Turing machine consists of a finite program, called the finite control, capable of manipulating a linear list of cells, called the

Reachability problem Given a Turing machine and a pair of states, decide if, starting from a configuration in the first state, the machine can reach a configuration in the second

We discuss a simple machine with 4 states and 3 symbols that has no periodic orbit and how that machine can be embedded into other ones to prove undecidability results on