• Aucun résultat trouvé

Syntactical complexity of periodic sets

N/A
N/A
Protected

Academic year: 2021

Partager "Syntactical complexity of periodic sets"

Copied!
29
0
0

Texte intégral

(1)

ÉmilieCharlier UniversitélibredeBruxelles

1stJointConferenceoftheBelgian,RoyalSpanishandLuxembourg MathematicalSocieties,June2012,Liège

(2)

0

1

0

0

1

0

1

0

1

0

0

0

13 8 5 3 2 1

1 0

2

1 0 1

4

1 0 0 1

6

1 0 0 0 0

8

1 0 0 1 0 10

1 0 1 0 1 12

. . .

(3)

The set

2N

ofeven integersis

F

-recognizable or

F

-automatic,i.e., the language

rep

F

(2N) = {ε, 10, 101, 1001, 10000, . . .}

isaccepted by somenite automaton.

Remark (in terms of the Chomsky hierarchy)

With respect tothe Zeckendorfsystem,any

F

-recognizable setcan be considered asaparticularlysimple set ofintegers.

(4)

0

1

0

I

F

n+2

= F

n+1

+ F

n

I

F

0

= 1

,

F

1

= 2

I

A

F

(5)

The

`

-bonacci numeration system

0

1

1

1

0

0

0

I

U

n+`

= U

n+`−1

+ U

n+`−2

+ · · · + U

n

I

U

i

= 2

i

,

i ∈ {0, . . . , ` − 1}

I

A

U

accepts allwordsthat do notcontain

1

`

(6)

U

-recognizability of arithmetic progressions

Proposition

Let

U = (U

i

)

i≥0

be anumeration systemand let

m, r ∈ N

.

If

N

is

U

-recognizable,then

m N +r

is

U

-recognizable and,given a DFAaccepting

rep

U

(N)

,aDFA accepting

rep

U

(m N +r)

can be obtained eectively.

(7)

U

-recognizability of

N

Is the set

N

U

-recognizable? Otherwisestated, isthe numeration language

rep

U

(N)

regular? Not necessarily:

Theorem(Shallit 1994)

Let

U

be aPNS. If

N

is

U

-recognizable,then

U

islinear,i.e.,it satises alinearrecurrence relationover

Z

.

The conditionis notsucient:

Example (

U

i

= (i + 1)

2

for all

i

∈ N

)

It is linear:

U

i+3

= 3U

i+2

− 3U

i+1

+ U

i

forall

i ∈ N

,but:

rep

U

(N) ∩ 10

10

= {10

a

10

b

: U

a+b+1

+ U

b

< U

a+b+2

}

= {10

a

10

b

: b

2

< 2a + 4}

Thus,

rep

(8)

What is the best automaton we can get?

0

1

2

3

0

1

0

1

0

1

0

1

0

1

3

0

1

0

1

0

1

DFAsacceptingthebinaryrepresentationsof

4N + 3

.

Question

The generalalgorithmdoesn't provide aminimalautomaton. What is the statecomplexity of

rep

(9)

Theorem(Krieger et al. 2009,Angrand-Sakarovitch 2010)

Let

m, r ∈ N

with

m ≥ 2

and

r < m

. If

rep

U

(N)

isaccepted by a

n

-stateDFA, thenthe minimal automaton of

rep

U

(

m

N

+ r)

has atmost

n

m

n

states.

NB: Thisresultremainstrue forthe larger classof abstract numeration systems.

(10)

Theorem(Alexeev 2004)

Let

b, m ≥ 2

. Let

N, M

besuch that

b

N

< m ≤ b

N +1

and

(m, 1) < (m, b) < · · · < (m, b

M

) = (m, b

M +1

)

.

The minimalautomaton recognizing

m

N

inbase

b

hasexactly

m

(

m

, b

N +1

)

+

inf

{N,M −1}

X

t=0

b

t

(

m

, b

t

)

states.

In particular, if

m

and

b

are coprime,then thisnumberisjust

m

. Further, if

m = b

n

(11)

Given any niteautomaton recognizingaset

X

of integerswritten in base

b

,itis decidablewhether

X

isultimatelyperiodic.

I

J.Honkala,Adecisionmethodfortherecognizabilityofsetsdenedbynumber systems,Theor. Inform. Appl. 20(1986).

I

A.Muchnick,ThedenablecriterionfordenabilityinPresburgerarithmetic anditsapplications,TCS290(2003).

I

J.Leroux,APolynomialTimePresburgerCriterionandSynthesisforNumber DecisionDiagrams,LICS2005(2005).

I

J.-P.Allouche,N.Rampersad,J.Shallit,Periodicity,repetitions,andorbitsof anautomaticsequence,TCS410(2009).

I

J.Bell,ÉC,A.Fraenkel,M.Rigo,Adecisionproblemforultimatelyperiodic setsinnon-standardnumerationsystems,IJAC19(2009).

I

F.Durand,DecidabilityoftheHD0Lultimateperiodicityproblem,arXiv(2011).

I

I.Mitrofanov,AproofforthedecidabilityofHD0Lultimateperiodicity,arXiv (2011).

(12)

Consider alinearnumerationsystem

U

such that

N

is

U

-recognizable.

How many states doesthetrim minimalautomaton

A

U,m

recognizing

m N

contain?

1. Give upper/lower bounds?

2. Study specialcases, e.g.,Zeckendorfnumeration system.

3. Get informationon the trim minimalautomaton

AU

recognizing

N

.

(13)

Theorem(C-Rampersad-Rigo-Waxweiler2011)

Let

U

be any numerationsystem (notnecessarily linear). The numberof statesof

A

U,

m

is atleast

| rep

U

(

m

)|

.

(14)

I

Let

U = (U

n

)

n≥0

bea linearnumeration system.

I

Let

k = kU,m

be the lengthofthe shortest linearrecurrence relationsatised by

(U

i

mod m)

i≥0

.

I

For

t ≥ 1

dene

H

t

:=

U

0

U

1

· · ·

U

t−1

U

1

U

2

· · ·

U

t

. . . . . . . . . . . .

U

t−1

U

t

· · ·

U

2t−2

.

I

For

m ≥ 2

,

kU,m

isalso the largest

t

suchthat

det H

t

6≡ 0

(15)

I

Let

SU,m

denote thenumberof

k

-tuples

b

in

{0, . . . , m − 1}

k

such thatthe system

H

k

x

≡ b

(mod m)

has atleast onesolution

x

= (x

1

, . . . , x

k

)

.

I

S

U,m

≤ m

k

(16)

Calculating

S

U,m

I

U

n+2

= 2U

n+1

+ U

n

,

(U

0

, U

1

) = (1, 3)

I

(U

n

)

n≥0

= 1, 3, 7, 17, 41, 99, 239, . . .

I

Consider the system

(

1 x

1

+ 3 x

2

≡ b

1

(mod 4)

3 x

1

+ 7 x

2

≡ b

2

(mod 4)

I

2x

1

≡ b

2

− b

1

(mod 4)

I

For each valueof

b

1

there areatmost

2

values for

b

2

.

I

(17)

Theorem

Let

m ≥ 2

bean integer. Let

U = (U

n

)

n≥0

be alinearnumeration system suchthat

(a)

N

is

U

-recognizable and

A

U

satises (H.1)and (H.2),

(b)

(U

n

mod m)

n≥0

ispurelyperiodic.

The numberof statesof

A

U,

m

fromwhichinnitelymany words areaccepted is

|C

U

| S

U,

m

.

(H.1)

A

U

has asinglestronglyconnected component

CU

.

(H.2) For allstates

p, q

in

C

U

with

p 6= q

,thereexists aword

x

pq

such that

δ

U

(p, x

pq

) ∈ C

U

and

δ

U

(q, x

pq

) 6∈ C

U

,orvice-versa.

(18)

Corollary

If

U

satisesthe conditions ofthe previoustheorem and

A

U

is strongly connected,then the numberof statesof

A

U,

m

is

(19)

Result for the

`

-bonacci system

0

1

1

1

0

0

0

Corollary

For

U

the

`

-bonacci system,the numberof states of

A

U,

m

is

`

m

`

(20)

0

1

0

0

1

0

1

0

1

0

0

0

13 8 5 3 2 1

1 0

2

1 0 1

4

1 0 0 1

6

1 0 0 0 0

8

1 0 0 1 0 10

1 0 1 0 1 12

. . .

(21)

I

Analyze the structure of

A

U

forsystemswith no dominant root.

I

Remove the assumptionthat

(U

n

mod m)

n≥0

ispurely periodic in the statecomplexity result.

I

(22)

Let

N

U

(m) ∈ {1, . . . , m}

denote the number ofvaluesthat are taken innitelyoftenby the sequence

(U

i

mod m)

i≥0

.

Example (Zeckendorf system)

(F

i

mod 4) = (1, 2, 3, 1, 0, 1, 1, 2, 3, . . .)

and

N

F

(4) = 4

.

(F

i

mod 11) = (1, 2, 3, 5, 8, 2, 10, 1, 0, 1, 1, 2, 3, . . .)

and

N

F

(11) = 7

.

Theorem(C-Rigo 2008)

Let

U = (U

i

)

i≥0

be anumeration systemsatisfying

lim

i→+∞

U

i+1

− U

i

= +∞

.

If

X ⊆ N

is anultimately periodic

U

-recognizable set ofperiod

p

, then any DFA accepting

rep

U

(X)

has atleast

N

U

(

p

)

states.

(23)

I

If

N

U

(m) → +∞

as

m → +∞

,thenwe obtain adecision procedure tothe periodicityproblem.

I

If

U

isaLNS satisfying

U

i+k

= a

1

U

i+k−1

+ · · · + a

k

U

i

, i ≥ 0,

with

a

k

= ±1,

then

lim

m→+∞

N

U

(m) = +∞

.

I

WorksfortheZeckendorf system.

I

Not trueforinteger baseb:

N (b

n

) = 1

(24)

I

Theformulaforthe state complexity of

m N

forthe Zeckendorf systemismuch simplerthan the formulafor integer base

b

systems.

I

Inthis point ofview, statecomplexityis not completely satisfying.

I

Hope: Findacomplexity that wouldhandle allthesesystems in akind ofuniformway.

(25)

I

Let

L

be alanguage over the nitealphabet

Σ

.

I

Myhill-Nerode equivalencerelationfor

L

:

u ∼

L

v

means that forall

y ∈ Σ

,

uy ∈ L ⇔ vy ∈ L

.

I

Leads tothe minimalautomaton of

L

:

|A

L

| = |Σ

/∼

L

|

is the state complexity of

L

.

I

Syntacticcongruence for

L

:

u ≡

L

v

means that forall

x, y ∈ Σ

,

xuy ∈ L ⇔ xvy ∈ L

.

I

Leads tothe syntactic monoidof

L

:

|H

L

| = |Σ

/≡

L

|

isthe

syntactic complexity of

L

.

Theorem

A language

L

is regularifand only if

Σ

/≡

(26)

The syntacticcomplexity of

X ⊆ N

isthe syntacticcomplexityof the language

0

rep

U

(X)

.

Let

ord

m

(b) = min{j ∈ N

0

: b

j

≡ 1 (mod m)}

.

Theorem(Rigo-Vandomme2011)

I

Let

m

, b ≥ 2

becoprimeintegers.

If

X ⊆ N

isperiodicof minimalperiod

m

,thenthe syntactic complexityof

X

is equal to

m

ord

m

(b)

.

(27)

I

Let

b ≥ 2

and

m

= b

n

with

n ≥ 1

.

(a) The syntacticcomplexityof

m

N

is equalto

2

n

+ 1

.

(b) If

X ⊆ N

isperiodicofminimalperiod

m

,thenthesyntactic complexityof

X

is

n

+ 1

.

I

Let

b ≥ 2

and

m

= b

n

q

with

n ≥ 1

and

(b, q) = 1

. Then the syntacticcomplexity of

m

N

is equalto

(28)

Theorem(Lacroix-Rampersad-Rigo-Vandomme,to appear)

Let

b ≥ 2

and

m

=

d

b

n

q

with

n ≥ 1

and

(b, q) = 1

and where

n

and

q

arechosen to be maximal.

If

X ⊆ N

is periodic ofminimalperiod

m

,thenthe syntactic complexity of

X

is

≥ max



q

ord

q

(b),

γ + 1

q

ord

q

(b)



,

where

γ → +∞

as

n

or

d

→ +∞

.

(29)

Theorem

The syntacticcomplexity of

m N

is

4m

2

p

F

(m) + 2

where

p

F

(m)

isthe minimalperiod of

(F

i

mod m)

i≥0

.

Further work forsyntactic complexity:

I

Tryto estimatethe syntacticcomplexity ofperiodic setsfora larger classof numerationsystems.

Syntactic complexity seemsto allowusto handleinteger basesand the Zeckendorf systematonce.

Références

Documents relatifs

Dans notre série de 45 patients, les symptômes ont été variables, marqués par la prédominance de l’épilepsie dans les localisations supra-tentorielles et

Consistency Given a model (a set of qualitative equation E ), the first question coming to mind is: is my model coherent?. A necessary condition for this is to have non

Since r is irreducible, it is either is a constant function of arity zero (in which case the reduction process ends), or it is decomposable into two coprime Boolean functions (using

The multiplicative structure (tensor product and its adjoint) can be defined in any locally cartesian closed category, whereas the additive (product and co- product) and exponential (

Then, we can build a UOTM which queries all these values (in time � i ≤N |f|(N), which is polynomial in the size of the inputs and the size of the oracles) and computes f on

Whereas there exist efficient algorithms for computing a NDD that represents the set defined by a given Presburger formula [14, 31, 4], the inverse problem of computing

If OMVD (resp. OMDD) satisfied ∧BC, we would have a poly- nomial algorithm deciding whether the conjunction of two OBDDs (the variable order being possibly different in each OBDD)