algorithmics in braid groups juan gonzález-meneses universidad de sevilla jornada temática...

49
Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona , October 22 – 23, 2010. Trenzas

Upload: leanna-gillian

Post on 31-Mar-2015

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Algorithmics in braid groups

Juan González-Meneses

Universidad de Sevilla

Jornada Temática Interdisciplinar de laRed Española de Topología

Barcelona , October 22 – 23, 2010.

Trenzas

Page 2: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Seville. June 13-17, 2011

www.imus.us.es/ACT/braids2011

Page 3: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computing in braid groups Garside structure

Juan González-Meneses

1

2345

E. Artin (1925)

Page 4: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computing in braid groups Garside structure

Juan González-Meneses

Monoid of positive braids:

Induces a lattice order:

Garside elementGarside element

Simple elements = Positive prefixes of

Page 5: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

1

G

P

Computing in braid groups Garside structure

Juan González-Meneses

Garside group

Dehornoy-Paris (1999)

And some otherproperties…

Page 6: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Lattice of simple elements in the braid group B4.

Juan González-Meneses

Computing in braid groups Garside structure

Page 7: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Word problem Left normal form

Juan González-Meneses

Garside (1969), Deligne (1972), Adyan (1984), Elrifai-Morton (1988), Thurston (1992).

Every braid x can be written in Left normal form:

Simple elements

Every product must be left-weighted.

Biggest possible simple element in any such writing of this product.Biggest possible simple element in any such writing of this product.

Artin (1925)

Page 8: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Word problem Left normal form

Juan González-Meneses

a s t

In general: Given simple elements

a b

simple simple

Not left-weighted

left-weighted

We call this procedure a local sliding applied to a b .

Page 9: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Word problem Left normal form

Juan González-Meneses

Computation of a left normal form, given a product of simple elements:

Apply all possible local slidings,until all consecutive factors are left weighted

Left normal form:

Maximal power of Minimal number of factors.

(canonical length)

Page 10: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Conjugacy problem

Juan González-Meneses

Charney: (1992) Artin-Tits groups of spherical type are biautomatic.

Garside (1969)

ElRifai-Morton (1988)

Birman-Ko-Lee (1998)

Franco-GM (2003)

Gebhardt (2005)

Birman-Gebhardt-GM (2008)

Gebhardt-GM (2009)

Page 11: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Conjugacy problem The main idea

Juan González-Meneses

Given an element x, compute the set of simplest conjugates of x.

(in some sense)

x and y are conjugate , their corresponding sets coincide. x and y are conjugate , their corresponding sets coincide.

Page 12: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Conjugacy problem Cyclic sliding

Juan González-Meneses

Can x be simplified by a conjugation?

For simplicity, we will assume that there is no power of :

Consecutive factors are left-weighted. What about x5 and x1 ?

Up to conjugacy, we can consider that x5 and x1 are consecutiveUp to conjugacy, we can consider that x5 and x1 are consecutive

Page 13: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Conjugacy problem Cyclic sliding

Juan González-Meneses

x2

x1 x 5

x4 x3

Page 14: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

x2

x1 x 5

x4 x3

t x2

x 5

x4 x3

Conjugacy problem Cyclic sliding

Juan González-Meneses

s Cyclic sliding:

x2

x 5 s

x4 x3

t

Page 15: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Conjugacy problem Cyclic sliding

Juan González-Meneses

x2

x 5 s

x4 x3

t Cyclic sliding: The resulting braid is notIn left normal form

The resulting braid is notIn left normal form

We compute its left normal form, and it may become simplerWe compute its left normal form, and it may become simpler

Iterate…

Page 16: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Conjugacy problem Sliding circuits

Juan González-Meneses

xsliding x1

sliding x2 x3

sliding …

Sliding circuit of x.Sliding circuit of x.

SC(x) Set of sliding circuits in the

Conjugacy class of x.

Set of sliding circuits in the

Conjugacy class of x.

Page 17: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computer experiments Random braids

Juan González-Meneses

The solution to the word problem is very efficient.

The solution to the conjugacy problem seems to be very efficient.

(taking braids at random)

At random?At random?

Many authors do computer experiments with “random braids”.

Page 18: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computer experiments Random braids

Quotes from papers on braid-cryptography:

Juan González-Meneses

“A” chooses a random secret braid…“A” chooses a random secret braid…

Ko, Lee, Cheon, Han, Kang, Park.

…we took 5000 random pairs of positive braids in Bn of length l (using Artin presentation)

…we took 5000 random pairs of positive braids in Bn of length l (using Artin presentation)

Franco, González-Meneses.

…the elements ai and bi are products of 10 random Artin generators.…the elements ai and bi are products of 10 random Artin generators.

Garber, Kaplan, Teicher, Tsaban,Vishne.

Random elements for these tests were obtained as follows. We choose independent random simple elements A1,A2, ... until len(A1/ Am) = r.

Random elements for these tests were obtained as follows. We choose independent random simple elements A1,A2, ... until len(A1/ Am) = r.

Gebhardt.

None of these procedures generates truly random braids!

Page 19: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computer experiments Random braids

We will focus on the positive braid monoid:

Juan González-Meneses

Length of a (positive) braid = Word lengthLength of a (positive) braid = Word length

We want to generate a random positive braid of length k.

What if we take the product of k randomly chosen generators? What if we take the product of k randomly chosen generators?

Page 20: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computer experiments Random braids

Juan González-Meneses

There is only one word representing the braid

There are two words representing the braid

The probability of obtaining is twice the probability of obtaining

This becomes more dramatic as n and k grow.

There are 16 words representing the braid

There is only one word representing the braid

The probability of obtaining is 16 times the probability of obtaining

Page 21: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids Lex-representative

Juan González-Meneses

How do we generate braids with the same probability?

Given a braid , its Lex-representative, w, is the (lexicographically) smallest word representing .

Given a braid , its Lex-representative, w, is the (lexicographically) smallest word representing .

{ Braids of length k } { Lex-representatives of length k }Bij.

Page 22: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids Braid tree

Juan González-Meneses

Example, in B4+ :

The situation can be described using a rooted tree.

Root.

The only braid of length 0 in B4+.

Page 23: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids Braid tree

Juan González-Meneses

Example, in B4+ :

The situation can be described using a rooted tree.

3 braids of length 1 in B4+.

Page 24: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids Braid tree

Juan González-Meneses

Example, in B4+ :

The situation can be described using a rooted tree.

8 braids of length 2 in B4+. The word is not there,

as the Lex-representativeof is .

Page 25: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids Braid tree

Juan González-Meneses

Example, in B4+ :

The situation can be described using a rooted tree.

19 braids of length 3 in B4+. There is no subword ,

neither , nor .

Page 26: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids Braid tree

Juan González-Meneses

Example, in B4+ :

The situation can be described using a rooted tree.

19 braids of length 3 in B4+. There is no subword ,

neither , nor .

Page 27: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids Generation procedure

Juan González-Meneses

To generate a random braid:

1) Compute the number of leaves of the tree: 19

2) Choose a random number between 1 and 19: 16

3) Find the braid corresponding to the 16th leaf: In polynomial time!In polynomial time!

Warning: the tree is exponentially big!Warning: the tree is exponentially big!

Page 28: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Bronfman method

Juan González-Meneses

x n,k = Number of braids in Bn+ of length k.

Recursive method to compute the growth function of Bn+

(want to compute this number)

Bronfman, 2001:

We wil use another method: simpler and faster.We wil use another method: simpler and faster.

Page 29: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

Example, in Bn+ :

= # Braids: Cannot start with

# Braids: Cannot start with

+ # Braids: # Braids:

Page 30: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

This yields an easy recursive formula:

+

+

+

-

-

=

Page 31: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

+

+

+

-

=

Example, in B4+ :

1 1 1 1

1233

Page 32: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

+

+

+

-

=

Example, in B4+ :

1 1 1 1

1233

2588

Page 33: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

+

+

+

-

=

Example, in B4+ :

1 1 1 1

1233

2588

4111919

Page 34: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

Example, in B4+ :

1 1 1 1

1233

2588

4111919

8244343+

+

+

-

=

Page 35: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

Example, in B4+ :

1 1 1 1

1233

2588

4111919

8244343

16519494+

+

+

-

=

Page 36: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

Example, in B4+ :

1 1 1 1

1233

2588

4111919

8244343

16519494

33108202202 …

What are we computing?What are we computing?

+

+

+

-

=

123

258

41119

Page 37: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Counting braids of given length Our method

Juan González-Meneses

123

258

41119

The first column of our table contains x n,1 , x n,2 , x n,3 ,…

Computing k rows, we obtain x n,k.Computing k rows, we obtain x n,k.

Page 38: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids Generation procedure

Juan González-Meneses

To generate a random braid:

1) Compute the number of leaves of the tree: 19

2) Choose a random number between 1 and 19: 16

3) Find the braid corresponding to the 16th leaf: NextNext

Page 39: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Finding the rth braid of length k Hanging leaves

Juan González-Meneses

Given a vertex w, suppose we can compute, in polynomial time,

the number of leaves hanging from w:

Given a vertex w, suppose we can compute, in polynomial time,

the number of leaves hanging from w:

Consider the graph of height k as before.

Page 40: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Finding the rth braid of length k The procedure

Juan González-Meneses

We want to compute the 16th braid:

The first letter is .The first letter is .

The second letter is .The second letter is .

The third letter is .The third letter is .

Page 41: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computing pending leaves Forbidden prefixes

Juan González-Meneses

Computing at most (n-2)k times the hanging leaves of a vertex,

one computes the rth braid of length k.

Computing at most (n-2)k times the hanging leaves of a vertex,

one computes the rth braid of length k.

How to compute the hanging leaves?

{ Leaves hanging from w } = { Lex-representatives starting with w }

Lemma: (GM, 2010) The number of leaves hanging from w is equal to

the number of braids, of length k-|w|, which cannot start with someforbidden prefixes.

Lemma: (GM, 2010) The number of leaves hanging from w is equal to

the number of braids, of length k-|w|, which cannot start with someforbidden prefixes.

Page 42: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computing pending leaves Forbidden prefixes

Juan González-Meneses

Example:

Forbidden prefixes for the word

Denote

Know how to compute

Page 43: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computing pending leaves Forbidden prefixes

Juan González-Meneses

By the inclusion-exclusion principle, we can compute

Page 44: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computing pending leaves Forbidden prefixes

Juan González-Meneses

By the inclusion-exclusion principle, we can compute

Page 45: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Computing pending leaves Forbidden prefixes

Juan González-Meneses

Good news: Every summand is equal to § xn,t, for some t· k.

These are the elements in the first column of our table.

But we can treat the same way all elements of the form a1Ç a2Ç Ç as

such that: ² Have the same length.

² Coincide in other technical questions, regarding permutation of strands.

Bad news: There are exponentially many summands.

Page 46: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random positive braids The result

Juan González-Meneses

As a consequence:

Theorem: (GM, 2010) There is a procedure to generate a random positive braid

in Bn+ of length k, whose time and space complexity is a polynomial in n and k.

Theorem: (GM, 2010) There is a procedure to generate a random positive braid

in Bn+ of length k, whose time and space complexity is a polynomial in n and k.

One can compute in ploynomial space and time with respect to n and k.

)

Page 47: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Every braid can be written, in a unique way, as:

Generating random braids From positive braids to braids

Juan González-Meneses

where is positive, and cannot start with .

There are xn,k positive braids of length k,

xn,k-|| of them can start with .1 1 1 1

1233

2588

4111919

8244343

16519494

33108202202Hence:

It is easy to show that

Page 48: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Generating random braids The result

Juan González-Meneses

Theorem: (GM, 2010) There is a partial algorithm to generate a random braid

in Bn, where | | = k, whose time and space complexity is a polynomial in n and k,

and whose rate of failure is < 2-n(n -1)/2.

Theorem: (GM, 2010) There is a partial algorithm to generate a random braid

in Bn, where | | = k, whose time and space complexity is a polynomial in n and k,

and whose rate of failure is < 2-n(n -1)/2.

Procedure: Choose a random .

Compute its left normal form.

If can start with , discard it and start again.

Probability: <

Page 49: Algorithmics in braid groups Juan González-Meneses Universidad de Sevilla Jornada Temática Interdisciplinar de la Red Española de Topología Barcelona,

Software to compute with braid groups

Juan González-Meneses

² www.personal.us.es/meneses

² www.indiana.edu/~knotinfo/

² GAP3

² MAGMA

(C++ source code)

(web applet)

(CHEVIE package)