37
Where innovation starts Laten zien dat het niet past/ How to show it doesn’t fit Stefan van Zwam Based on joint work with Rhiannon Hall and Dillon Mayhew Nederlands Mathematisch Congres, April 15, 2009

Laten zien dat het niet past/ How to show it doesn’t fitsvanzwam/talks/20090416.pdf · Laten zien dat het niet past/ How to show it doesn’t fit ... 1900–1936: Dedekind, Birkhoff,

  • Upload
    ngomien

  • View
    214

  • Download
    0

Embed Size (px)

Citation preview

Where innovation starts

Laten zien dat het niet past/How to show it doesn’t fit

Stefan van Zwam

Based on joint work with Rhiannon Hall and Dillon Mayhew

Nederlands Mathematisch Congres, April 15, 2009

2/31

/ department of mathematics and computer science

Matroid theory

3/31

/ department of mathematics and computer science

Linearly independent vectors in Rn

Example

4/31

/ department of mathematics and computer science

Linearly independent vectors in Rn

Example

5/31

/ department of mathematics and computer science

Linearly independent vectors in Rn

Example

6/31

/ department of mathematics and computer science

Lemma. Given

E: finite set of vectors

I: collection of linearly independent subsets

then

•∅ ∈ I• J ∈ I and ⊆ J, then ∈ I• , J ∈ I and || < |J|, then

∃e ∈ J \ such that ∪ {e} ∈ I

Matroid axioms

7/31

/ department of mathematics and computer science

Definition. Given

E: finite set

I: collection of subsets

such that

•∅ ∈ I• J ∈ I and ⊆ J, then ∈ I• , J ∈ I and || < |J|, then

∃e ∈ J \ such that ∪ {e} ∈ I

Then M = (E,I) is a matroid.

Matroid axioms

8/31

/ department of mathematics and computer science

1899: Hilbert, Bernays: Plane geometry

1900–1936: Dedekind, Birkhoff, MacLane:Semimodular lattices

1910–1937: Steinitz, Van der Waerden:Algebraic dependence

1936: Nakasawa: Projective geometry

1935: Whitney: Lin. dependence, duality, graphs

1942: Rado: Transversals (matching theory)

1958: Tutte: Connectivity, minors, . . .

1971: Edmonds: Greedy algorithm

Rota, Brylawski, Seymour, . . .

Matroids everywhere!

9/31

/ department of mathematics and computer science

The representation problem

Problem. Is there a map

E→ Fn

preserving the dependencies of M = (E,I)?

Matroid representation

10/31

/ department of mathematics and computer science

Example: the Fano matroid

100

001

010

• E = { points }

• I = {X ⊆ E in general position }

Matroid representation

11/31

/ department of mathematics and computer science

Example: the Fano matroid

100

001

010

• E = { points }

• I = {X ⊆ E in general position }

Matroid representation

12/31

/ department of mathematics and computer science

Example: the Fano matroid

100

001

010

• E = { points }

• I = {X ⊆ E in general position }

Matroid representation

13/31

/ department of mathematics and computer science

Example: the Fano matroid

100

001

010

• E = { points }

• I = {X ⊆ E in general position }

Matroid representation

14/31

/ department of mathematics and computer science

Example: the Fano matroid

100

001

010

101

011

111

110

• E = { points }

• I = {X ⊆ E in general position }

Matroid representation

15/31

/ department of mathematics and computer science

Example: the Fano matroid

100

001

010

101

011

111

110

110

−101

−011

= −002

Matroid representation

16/31

/ department of mathematics and computer science

How to show it doesn’t fit?

Problem. Is there a dependency-preserving map

E(M)→GF(2)

F

...

?

Matroid representation

16/31

/ department of mathematics and computer science

How to show it doesn’t fit?

Problem. Is there a dependency-preserving map

E(M)→GF(2)

F

...

?

• “Yes” certified by vectors {1, . . . , n}

Matroid representation

16/31

/ department of mathematics and computer science

How to show it doesn’t fit?

Problem. Is there a dependency-preserving map

E(M)→GF(2)

F

...

?

• “Yes” certified by vectors {1, . . . , n}

• How to certify “no”?

Matroid representation

17/31

/ department of mathematics and computer science

Reducing a set of vectors: deletion

Matroid minors

18/31

/ department of mathematics and computer science

Reducing a set of vectors: deletion

Matroid minors

19/31

/ department of mathematics and computer science

Reducing a set of vectors: projection

Matroid minors

20/31

/ department of mathematics and computer science

Reducing a set of vectors: projection

Matroid minors

21/31

/ department of mathematics and computer science

Reducing a set of vectors: projection

Matroid minors

22/31

/ department of mathematics and computer science

Abstract definition• Deletion: M\e := (E \ {e},{ ∈ I : e 6∈ })• Contraction: M/e := (E \ {e},{ : ∪ {e} ∈ I})• Minors: Obtained from sequence of such steps

Matroid minors

22/31

/ department of mathematics and computer science

Abstract definition• Deletion: M\e := (E \ {e},{ ∈ I : e 6∈ })• Contraction: M/e := (E \ {e},{ : ∪ {e} ∈ I})• Minors: Obtained from sequence of such steps

– Generate partial order– Preserve representability

Matroid minors

23/31

/ department of mathematics and computer science

Excluded minors

Matroid minors

24/31

/ department of mathematics and computer science

That’s how!

Problem:Is there a dependency-preserving map

¨M : E(M)→

GF(2)

F

...

«

• How to certify the answer is “no”?

• By reducing to an excluded minor!

Matroid representation

24/31

/ department of mathematics and computer science

That’s how!

Problem:Is there a dependency-preserving map

¨M : E(M)→

GF(2)

F

...

«

• How to certify the answer is “no”?

• By reducing to an excluded minor!

• Rota’s Conjecture: finitely many

Matroid representation

25/31

/ department of mathematics and computer science

Rota’s Conjecture

Conjecture (Rota 1971): F finite, then ∃k = k(F) :exactly k excluded minors for

¨M : E(M)→

GF(2)

F

...

«

Matroid representation

25/31

/ department of mathematics and computer science

Rota’s Conjecture

Conjecture (Rota 1971): F finite, then ∃k = k(F) :exactly k excluded minors for

¨M : E(M)→

GF(2)

F

...

«

• Proven for F ∈ {GF(2),GF(3),GF(4)}

Matroid representation

26/31

/ department of mathematics and computer science

Regular matroids

Theorem (Tutte 1958):Exactly 3 excluded minors for

¨M : E(M)→

GF(2)GF(3)GF(4)GF(5)GF(7)...

«

Matroid representation

27/31

/ department of mathematics and computer science

Near-regular matroids

Theorem (Hall, Mayhew, vZ 2009):Exactly 10 excluded minors for

¨M : E(M)→

GF(2)GF(3)GF(4)GF(5)GF(7)...

«

Matroid representation

28/31

/ department of mathematics and computer science

Excluded minors for near-regular¨, , ,

� �∗,

,� �∗

, ,

,� �∗

,� �ΔY«

Matroid representation

29/31

/ department of mathematics and computer science

Near-regular matroids

Theorem (Hall, Mayhew, vZ 2009):Exactly 10 excluded minors for

¨M : E(M)→

GF(2)GF(3)GF(4)GF(5)GF(7)...

«

Matroid representation

30/31

/ department of mathematics and computer science

Near-regular matroids

Theorem (Hall, Mayhew, vZ 2009):Exactly 10 excluded minors for

¨M : E(M)→

GF(2)

P

...

«

Matroid representation

30/31

/ department of mathematics and computer science

Near-regular matroids

Theorem (Hall, Mayhew, vZ 2009):Exactly 10 excluded minors for

¨M : E(M)→

GF(2)

P

...

«Why care?• Complexity of algorithms

• Structure of ternary matroids

• Sharpening the knives for Rota’s Conjecture

Matroid representation

31/31

/ department of mathematics and computer science

Summary

• Matroid axioms abstract linear dependence

• Matroids occur everywhere

• Representation problem

• Excluded minors

• Near-regular matroids

Thank you for listening

Preprint at http://www.win.tue.nl/~svzwam/

The end