On the Banach-Saks property and convex hulls J. Lopez-Abad - - PowerPoint PPT Presentation

on the banach saks property and convex hulls
SMART_READER_LITE
LIVE PREVIEW

On the Banach-Saks property and convex hulls J. Lopez-Abad - - PowerPoint PPT Presentation

On the Banach-Saks property and convex hulls On the Banach-Saks property and convex hulls J. Lopez-Abad Instituto de Ciencias Matem aticas CSIC, Madrid This is a joint work with C. Ruiz-Bermejo and P . Tradacete (Madrid) Trends in Set


slide-1
SLIDE 1

On the Banach-Saks property and convex hulls

On the Banach-Saks property and convex hulls

  • J. Lopez-Abad

Instituto de Ciencias Matem´ aticas CSIC, Madrid

This is a joint work with C. Ruiz-Bermejo and P . Tradacete (Madrid)

Trends in Set Theory 2012

slide-2
SLIDE 2

On the Banach-Saks property and convex hulls Introduction

A classical theorem of Mazur asserts that the convex hull of a compact set in a Banach space is again relatively compact. Indeed, Krein- ˇ Smulian’s Theorem states that the same holds even for weakly compact sets. There is a well-known property lying between these two main kinds of compactness: Definition A subset A ⊆ X of a Banach space is called Banach-Saks if every sequence in A has a Ces` aro convergent subsequence, i.e. every sequence (xn)n in A has a subsequence (xnk)k such that the sequence of means (1 k

k

  • i=1

xni)k converges in norm.

slide-3
SLIDE 3

On the Banach-Saks property and convex hulls Introduction

A classical theorem of Mazur asserts that the convex hull of a compact set in a Banach space is again relatively compact. Indeed, Krein- ˇ Smulian’s Theorem states that the same holds even for weakly compact sets. There is a well-known property lying between these two main kinds of compactness: Definition A subset A ⊆ X of a Banach space is called Banach-Saks if every sequence in A has a Ces` aro convergent subsequence, i.e. every sequence (xn)n in A has a subsequence (xnk)k such that the sequence of means (1 k

k

  • i=1

xni)k converges in norm.

slide-4
SLIDE 4

On the Banach-Saks property and convex hulls Introduction

A classical theorem of Mazur asserts that the convex hull of a compact set in a Banach space is again relatively compact. Indeed, Krein- ˇ Smulian’s Theorem states that the same holds even for weakly compact sets. There is a well-known property lying between these two main kinds of compactness: Definition A subset A ⊆ X of a Banach space is called Banach-Saks if every sequence in A has a Ces` aro convergent subsequence, i.e. every sequence (xn)n in A has a subsequence (xnk)k such that the sequence of means (1 k

k

  • i=1

xni)k converges in norm.

slide-5
SLIDE 5

On the Banach-Saks property and convex hulls Introduction

A classical theorem of Mazur asserts that the convex hull of a compact set in a Banach space is again relatively compact. Indeed, Krein- ˇ Smulian’s Theorem states that the same holds even for weakly compact sets. There is a well-known property lying between these two main kinds of compactness: Definition A subset A ⊆ X of a Banach space is called Banach-Saks if every sequence in A has a Ces` aro convergent subsequence, i.e. every sequence (xn)n in A has a subsequence (xnk)k such that the sequence of means (1 k

k

  • i=1

xni)k converges in norm.

slide-6
SLIDE 6

On the Banach-Saks property and convex hulls Introduction

A classical theorem of Mazur asserts that the convex hull of a compact set in a Banach space is again relatively compact. Indeed, Krein- ˇ Smulian’s Theorem states that the same holds even for weakly compact sets. There is a well-known property lying between these two main kinds of compactness: Definition A subset A ⊆ X of a Banach space is called Banach-Saks if every sequence in A has a Ces` aro convergent subsequence, i.e. every sequence (xn)n in A has a subsequence (xnk)k such that the sequence of means (1 k

k

  • i=1

xni)k converges in norm.

slide-7
SLIDE 7

On the Banach-Saks property and convex hulls Introduction

A space has the Banach-Saks property when its unit ball is a Banach-Saks set. Examples of Banach-Saks sets are

1 the unit balls of ℓp’s, 1 < p < ∞ 2 the unit basis of c0.

Typical example of a weakly-null sequence which is not a Banach-Saks set is the unit basis of the Shreier space.

slide-8
SLIDE 8

On the Banach-Saks property and convex hulls Introduction

A space has the Banach-Saks property when its unit ball is a Banach-Saks set. Examples of Banach-Saks sets are

1 the unit balls of ℓp’s, 1 < p < ∞ 2 the unit basis of c0.

Typical example of a weakly-null sequence which is not a Banach-Saks set is the unit basis of the Shreier space.

slide-9
SLIDE 9

On the Banach-Saks property and convex hulls Introduction

A space has the Banach-Saks property when its unit ball is a Banach-Saks set. Examples of Banach-Saks sets are

1 the unit balls of ℓp’s, 1 < p < ∞ 2 the unit basis of c0.

Typical example of a weakly-null sequence which is not a Banach-Saks set is the unit basis of the Shreier space.

slide-10
SLIDE 10

On the Banach-Saks property and convex hulls Introduction

A space has the Banach-Saks property when its unit ball is a Banach-Saks set. Examples of Banach-Saks sets are

1 the unit balls of ℓp’s, 1 < p < ∞ 2 the unit basis of c0.

Typical example of a weakly-null sequence which is not a Banach-Saks set is the unit basis of the Shreier space.

slide-11
SLIDE 11

On the Banach-Saks property and convex hulls Introduction

A space has the Banach-Saks property when its unit ball is a Banach-Saks set. Examples of Banach-Saks sets are

1 the unit balls of ℓp’s, 1 < p < ∞ 2 the unit basis of c0.

Typical example of a weakly-null sequence which is not a Banach-Saks set is the unit basis of the Shreier space.

slide-12
SLIDE 12

On the Banach-Saks property and convex hulls Introduction

Question Is the convex hull of a Banach-Saks set again Banach-Saks? By Ramsey-like methods, we show that the answer is No: Theorem (LA-Ruiz-Tradacete) There is a family F of finite subsets of N such that the unit basis of a Shreier-like space XF is a Banach-Saks set, but its convex hull is not. On the opposite direction we prove that Theorem (LA-Ruiz-Tradacete) Suppose that F is a “classical” family (i.e. if F is a generalized Schreier family), then the convex hull of a Banach-Saks subset

  • f XF is also Banach-Saks.
slide-13
SLIDE 13

On the Banach-Saks property and convex hulls Introduction

Question Is the convex hull of a Banach-Saks set again Banach-Saks? By Ramsey-like methods, we show that the answer is No: Theorem (LA-Ruiz-Tradacete) There is a family F of finite subsets of N such that the unit basis of a Shreier-like space XF is a Banach-Saks set, but its convex hull is not. On the opposite direction we prove that Theorem (LA-Ruiz-Tradacete) Suppose that F is a “classical” family (i.e. if F is a generalized Schreier family), then the convex hull of a Banach-Saks subset

  • f XF is also Banach-Saks.
slide-14
SLIDE 14

On the Banach-Saks property and convex hulls Introduction

Question Is the convex hull of a Banach-Saks set again Banach-Saks? By Ramsey-like methods, we show that the answer is No: Theorem (LA-Ruiz-Tradacete) There is a family F of finite subsets of N such that the unit basis of a Shreier-like space XF is a Banach-Saks set, but its convex hull is not. On the opposite direction we prove that Theorem (LA-Ruiz-Tradacete) Suppose that F is a “classical” family (i.e. if F is a generalized Schreier family), then the convex hull of a Banach-Saks subset

  • f XF is also Banach-Saks.
slide-15
SLIDE 15

On the Banach-Saks property and convex hulls Introduction

Question Is the convex hull of a Banach-Saks set again Banach-Saks? By Ramsey-like methods, we show that the answer is No: Theorem (LA-Ruiz-Tradacete) There is a family F of finite subsets of N such that the unit basis of a Shreier-like space XF is a Banach-Saks set, but its convex hull is not. On the opposite direction we prove that Theorem (LA-Ruiz-Tradacete) Suppose that F is a “classical” family (i.e. if F is a generalized Schreier family), then the convex hull of a Banach-Saks subset

  • f XF is also Banach-Saks.
slide-16
SLIDE 16

On the Banach-Saks property and convex hulls Introduction

Question Is the convex hull of a Banach-Saks set again Banach-Saks? By Ramsey-like methods, we show that the answer is No: Theorem (LA-Ruiz-Tradacete) There is a family F of finite subsets of N such that the unit basis of a Shreier-like space XF is a Banach-Saks set, but its convex hull is not. On the opposite direction we prove that Theorem (LA-Ruiz-Tradacete) Suppose that F is a “classical” family (i.e. if F is a generalized Schreier family), then the convex hull of a Banach-Saks subset

  • f XF is also Banach-Saks.
slide-17
SLIDE 17

On the Banach-Saks property and convex hulls Schreier families

Definition Let F be a family on N, i.e. a collection of (finite) subsets of N. Given x ∈ c00(N) we define xF := max{x∞, sup

s∈F

  • k∈s

|(x)k|}. The Shreier-like space XF is the completion of (c00(N), · F). It is easy to see that the unit basis (un)n of c00(N) is a 1-unconditional Schauder basis of XF. The non-trivial spaces are coming from pre-compact families, i.e. such that F ⊆ FIN.

slide-18
SLIDE 18

On the Banach-Saks property and convex hulls Schreier families

Definition Let F be a family on N, i.e. a collection of (finite) subsets of N. Given x ∈ c00(N) we define xF := max{x∞, sup

s∈F

  • k∈s

|(x)k|}. The Shreier-like space XF is the completion of (c00(N), · F). It is easy to see that the unit basis (un)n of c00(N) is a 1-unconditional Schauder basis of XF. The non-trivial spaces are coming from pre-compact families, i.e. such that F ⊆ FIN.

slide-19
SLIDE 19

On the Banach-Saks property and convex hulls Schreier families

Definition Let F be a family on N, i.e. a collection of (finite) subsets of N. Given x ∈ c00(N) we define xF := max{x∞, sup

s∈F

  • k∈s

|(x)k|}. The Shreier-like space XF is the completion of (c00(N), · F). It is easy to see that the unit basis (un)n of c00(N) is a 1-unconditional Schauder basis of XF. The non-trivial spaces are coming from pre-compact families, i.e. such that F ⊆ FIN.

slide-20
SLIDE 20

On the Banach-Saks property and convex hulls Schreier families

Definition Let F be a family on N, i.e. a collection of (finite) subsets of N. Given x ∈ c00(N) we define xF := max{x∞, sup

s∈F

  • k∈s

|(x)k|}. The Shreier-like space XF is the completion of (c00(N), · F). It is easy to see that the unit basis (un)n of c00(N) is a 1-unconditional Schauder basis of XF. The non-trivial spaces are coming from pre-compact families, i.e. such that F ⊆ FIN.

slide-21
SLIDE 21

On the Banach-Saks property and convex hulls Schreier families

Definition Let F be a family on N, i.e. a collection of (finite) subsets of N. Given x ∈ c00(N) we define xF := max{x∞, sup

s∈F

  • k∈s

|(x)k|}. The Shreier-like space XF is the completion of (c00(N), · F). It is easy to see that the unit basis (un)n of c00(N) is a 1-unconditional Schauder basis of XF. The non-trivial spaces are coming from pre-compact families, i.e. such that F ⊆ FIN.

slide-22
SLIDE 22

On the Banach-Saks property and convex hulls Schreier families

Let S = {s ∈ FIN : |s| ≤ min s + 1} be the Schreier family. Then the unit basis {un}n of XS is a non-Banach-Saks:The main reason is that S is large in N: For every n ∈ N and M ⊆ N there is s ∈ S ∩ [M]n. However, Proposition (M. Gonz´ alez, and J. Guti´ errez) The convex hull of a Banach-Saks subset of XS is a Banach-Saks set.

slide-23
SLIDE 23

On the Banach-Saks property and convex hulls Schreier families

Let S = {s ∈ FIN : |s| ≤ min s + 1} be the Schreier family. Then the unit basis {un}n of XS is a non-Banach-Saks:The main reason is that S is large in N: For every n ∈ N and M ⊆ N there is s ∈ S ∩ [M]n. However, Proposition (M. Gonz´ alez, and J. Guti´ errez) The convex hull of a Banach-Saks subset of XS is a Banach-Saks set.

slide-24
SLIDE 24

On the Banach-Saks property and convex hulls Schreier families

Let S = {s ∈ FIN : |s| ≤ min s + 1} be the Schreier family. Then the unit basis {un}n of XS is a non-Banach-Saks:The main reason is that S is large in N: For every n ∈ N and M ⊆ N there is s ∈ S ∩ [M]n. However, Proposition (M. Gonz´ alez, and J. Guti´ errez) The convex hull of a Banach-Saks subset of XS is a Banach-Saks set.

slide-25
SLIDE 25

On the Banach-Saks property and convex hulls Schreier families

Let S = {s ∈ FIN : |s| ≤ min s + 1} be the Schreier family. Then the unit basis {un}n of XS is a non-Banach-Saks:The main reason is that S is large in N: For every n ∈ N and M ⊆ N there is s ∈ S ∩ [M]n. However, Proposition (M. Gonz´ alez, and J. Guti´ errez) The convex hull of a Banach-Saks subset of XS is a Banach-Saks set.

slide-26
SLIDE 26

On the Banach-Saks property and convex hulls Schreier families

Let S = {s ∈ FIN : |s| ≤ min s + 1} be the Schreier family. Then the unit basis {un}n of XS is a non-Banach-Saks:The main reason is that S is large in N: For every n ∈ N and M ⊆ N there is s ∈ S ∩ [M]n. However, Proposition (M. Gonz´ alez, and J. Guti´ errez) The convex hull of a Banach-Saks subset of XS is a Banach-Saks set.

slide-27
SLIDE 27

On the Banach-Saks property and convex hulls Schreier families

Given two families F and G on N, let F ⊕ G :={s ∪ t : s ∈ G, t ∈ F and s < t} F ⊗ G :={

  • i

si : {si}i ⊆ F, si < sj for i < j, and {min si}i ∈ G}. Definition For each α < ω1, α limit, we fix a strictly increasing sequence (β(α)

n )n such that supn β(α) n

= α. We define now (a) S0 := [N]≤1. (b) Sα+1 = Sα ⊗ S. (c) Sα :=

n∈N Sβ(α)

n

↾ N/n. Then each Sα is a compact, hereditary and spreading family.

slide-28
SLIDE 28

On the Banach-Saks property and convex hulls Schreier families

Given two families F and G on N, let F ⊕ G :={s ∪ t : s ∈ G, t ∈ F and s < t} F ⊗ G :={

  • i

si : {si}i ⊆ F, si < sj for i < j, and {min si}i ∈ G}. Definition For each α < ω1, α limit, we fix a strictly increasing sequence (β(α)

n )n such that supn β(α) n

= α. We define now (a) S0 := [N]≤1. (b) Sα+1 = Sα ⊗ S. (c) Sα :=

n∈N Sβ(α)

n

↾ N/n. Then each Sα is a compact, hereditary and spreading family.

slide-29
SLIDE 29

On the Banach-Saks property and convex hulls Schreier families

Theorem For every pre-compact family F on N there are α < ω1, n ∈ N and M ⊆ N such that Sα[M] ⊕ [M]≤n ⊆ F[M] ⊆ Sα[M] ⊕ [M]≤n+1, where F[M] := {s ∩ M : s ∈ F}.

slide-30
SLIDE 30

On the Banach-Saks property and convex hulls Schreier families

Theorem (LA-Ruiz-Tradacete) The convex hull of a Banach-Saks subset of XSα is also a Banach-Saks set.

slide-31
SLIDE 31

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Some definitions. Definition

1 By a family on an infinite subset M ⊆ N we mean a

collection of finite subsets of M.

2 A family F on M is called large in N ⊆ M when for every

infinite subset P ⊆ N and every n ∈ N there is some s ∈ F such that |s ∩ P| ≥ n.

3 Given a partition (In)n of N, a transversal (w.r.t. (In)n) is a

subset T of N such that |T ∩ In| ≤ 1 for every n ∈ N.

slide-32
SLIDE 32

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Some definitions. Definition

1 By a family on an infinite subset M ⊆ N we mean a

collection of finite subsets of M.

2 A family F on M is called large in N ⊆ M when for every

infinite subset P ⊆ N and every n ∈ N there is some s ∈ F such that |s ∩ P| ≥ n.

3 Given a partition (In)n of N, a transversal (w.r.t. (In)n) is a

subset T of N such that |T ∩ In| ≤ 1 for every n ∈ N.

slide-33
SLIDE 33

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Some definitions. Definition

1 By a family on an infinite subset M ⊆ N we mean a

collection of finite subsets of M.

2 A family F on M is called large in N ⊆ M when for every

infinite subset P ⊆ N and every n ∈ N there is some s ∈ F such that |s ∩ P| ≥ n.

3 Given a partition (In)n of N, a transversal (w.r.t. (In)n) is a

subset T of N such that |T ∩ In| ≤ 1 for every n ∈ N.

slide-34
SLIDE 34

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Some definitions. Definition

1 By a family on an infinite subset M ⊆ N we mean a

collection of finite subsets of M.

2 A family F on M is called large in N ⊆ M when for every

infinite subset P ⊆ N and every n ∈ N there is some s ∈ F such that |s ∩ P| ≥ n.

3 Given a partition (In)n of N, a transversal (w.r.t. (In)n) is a

subset T of N such that |T ∩ In| ≤ 1 for every n ∈ N.

slide-35
SLIDE 35

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Definition A T-family is a family F of finite subsets of N such that there is a partition

n In of N in finite pieces In and for each n probability

measures µn on P(In) (i.e. a convex combination (λ(n)

k )k∈In)

with the following properties: (a) There is some ε > 0 and some M ⊆ N such that the set G(F, ε) := {t ⊆ N : ∃s ∈ F ∀n ∈ t µn(s ∩ In) ≥ ε } is large in M. (b) F[T] := {s ∩ T : s ∈ F} is not large in T for every transversal T ⊆ I.

slide-36
SLIDE 36

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Definition A T-family is a family F of finite subsets of N such that there is a partition

n In of N in finite pieces In and for each n probability

measures µn on P(In) (i.e. a convex combination (λ(n)

k )k∈In)

with the following properties: (a) There is some ε > 0 and some M ⊆ N such that the set G(F, ε) := {t ⊆ N : ∃s ∈ F ∀n ∈ t µn(s ∩ In) ≥ ε } is large in M. (b) F[T] := {s ∩ T : s ∈ F} is not large in T for every transversal T ⊆ I.

slide-37
SLIDE 37

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Definition A T-family is a family F of finite subsets of N such that there is a partition

n In of N in finite pieces In and for each n probability

measures µn on P(In) (i.e. a convex combination (λ(n)

k )k∈In)

with the following properties: (a) There is some ε > 0 and some M ⊆ N such that the set G(F, ε) := {t ⊆ N : ∃s ∈ F ∀n ∈ t µn(s ∩ In) ≥ ε } is large in M. (b) F[T] := {s ∩ T : s ∈ F} is not large in T for every transversal T ⊆ I.

slide-38
SLIDE 38

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Definition A T-family is a family F of finite subsets of N such that there is a partition

n In of N in finite pieces In and for each n probability

measures µn on P(In) (i.e. a convex combination (λ(n)

k )k∈In)

with the following properties: (a) There is some ε > 0 and some M ⊆ N such that the set G(F, ε) := {t ⊆ N : ∃s ∈ F ∀n ∈ t µn(s ∩ In) ≥ ε } is large in M. (b) F[T] := {s ∩ T : s ∈ F} is not large in T for every transversal T ⊆ I.

slide-39
SLIDE 39

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Theorem (LA-Ruiz-Tradacete) TFAE:

1 The convex hull of every weakly-null Banach-Saks set is

Banach-Saks.

2 There are no T-families.

Theorem (LA-Ruiz-Tradacete) There is a T-family (where indeed the measures µn on In are the counting measures on In).

slide-40
SLIDE 40

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Theorem (LA-Ruiz-Tradacete) TFAE:

1 The convex hull of every weakly-null Banach-Saks set is

Banach-Saks.

2 There are no T-families.

Theorem (LA-Ruiz-Tradacete) There is a T-family (where indeed the measures µn on In are the counting measures on In).

slide-41
SLIDE 41

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Theorem (LA-Ruiz-Tradacete) TFAE:

1 The convex hull of every weakly-null Banach-Saks set is

Banach-Saks.

2 There are no T-families.

Theorem (LA-Ruiz-Tradacete) There is a T-family (where indeed the measures µn on In are the counting measures on In).

slide-42
SLIDE 42

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Theorem (LA-Ruiz-Tradacete) TFAE:

1 The convex hull of every weakly-null Banach-Saks set is

Banach-Saks.

2 There are no T-families.

Theorem (LA-Ruiz-Tradacete) There is a T-family (where indeed the measures µn on In are the counting measures on In).

slide-43
SLIDE 43

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Theorem (LA-Ruiz-Tradacete) TFAE:

1 The convex hull of every weakly-null Banach-Saks set is

Banach-Saks.

2 There are no T-families.

Theorem (LA-Ruiz-Tradacete) There is a T-family (where indeed the measures µn on In are the counting measures on In).

slide-44
SLIDE 44

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Proposition Suppose that F is a T-family on N with respect to (In)n and (µn)n. Then

1 Every subsequence of the unit basis ¯

u = (un)n of XF has a further subsequence equivalent to the unit basis of c0; hence {un}n is a Banach-Saks set.

2 the set {µn ∗ ¯

x}n ⊆ conv({un}n) is not Banach-Saks.

slide-45
SLIDE 45

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Proposition Suppose that F is a T-family on N with respect to (In)n and (µn)n. Then

1 Every subsequence of the unit basis ¯

u = (un)n of XF has a further subsequence equivalent to the unit basis of c0; hence {un}n is a Banach-Saks set.

2 the set {µn ∗ ¯

x}n ⊆ conv({un}n) is not Banach-Saks.

slide-46
SLIDE 46

On the Banach-Saks property and convex hulls Combinatorial reformulation of the general problem

Proposition Suppose that F is a T-family on N with respect to (In)n and (µn)n. Then

1 Every subsequence of the unit basis ¯

u = (un)n of XF has a further subsequence equivalent to the unit basis of c0; hence {un}n is a Banach-Saks set.

2 the set {µn ∗ ¯

x}n ⊆ conv({un}n) is not Banach-Saks.

slide-47
SLIDE 47

On the Banach-Saks property and convex hulls Existence of T-families

Recall the following well-known fact: Theorem (Gillis) For every ε > 0, every δ > 0 and every m ∈ N there is n := n(ε, δ, m) such that for every probability space (Ω, F, µ) and every sequence (Ai)i<n such that µ(Ai) ≥ ε for all i < n, there is s ∈ [n]m such that µ(

  • i∈s

Ai) ≥ (1 − δ)εm.

slide-48
SLIDE 48

On the Banach-Saks property and convex hulls Existence of T-families

Recall the following well-known fact: Theorem (Gillis) For every ε > 0, every δ > 0 and every m ∈ N there is n := n(ε, δ, m) such that for every probability space (Ω, F, µ) and every sequence (Ai)i<n such that µ(Ai) ≥ ε for all i < n, there is s ∈ [n]m such that µ(

  • i∈s

Ai) ≥ (1 − δ)εm.

slide-49
SLIDE 49

On the Banach-Saks property and convex hulls Existence of T-families

Question Given ε > 0 and m ∈ N, does there exist n := n(ε, m) such that whenever (Ω, F, µ) is a probability space and (A{i,j}){i,j}∈[n]2 are such that µ(A{i,j}) > ε for all {i, j} ∈ [n]2, then there is s ∈ [n]m such that

{i,j}∈[s]2 A{i,j} = ∅?

NO: Example by Erd¨

  • s and Hajnal: Fix arbitrary n, r > 0. For

each i, j ∈ [n]2, let A{i,j} := {(ak)k<n ∈ r n : ai = aj}. Consider r n with its counting probability measure, µ(s) = |s|/r n for s ⊆ r n. Then

1 µ(A{i,j}) ≥ 1 − 1/r, and 2 {i,j}∈[s]2 A{i,j} = ∅ for every s ⊆ n with |s| ≥ r + 1.

slide-50
SLIDE 50

On the Banach-Saks property and convex hulls Existence of T-families

Question Given ε > 0 and m ∈ N, does there exist n := n(ε, m) such that whenever (Ω, F, µ) is a probability space and (A{i,j}){i,j}∈[n]2 are such that µ(A{i,j}) > ε for all {i, j} ∈ [n]2, then there is s ∈ [n]m such that

{i,j}∈[s]2 A{i,j} = ∅?

NO: Example by Erd¨

  • s and Hajnal: Fix arbitrary n, r > 0. For

each i, j ∈ [n]2, let A{i,j} := {(ak)k<n ∈ r n : ai = aj}. Consider r n with its counting probability measure, µ(s) = |s|/r n for s ⊆ r n. Then

1 µ(A{i,j}) ≥ 1 − 1/r, and 2 {i,j}∈[s]2 A{i,j} = ∅ for every s ⊆ n with |s| ≥ r + 1.

slide-51
SLIDE 51

On the Banach-Saks property and convex hulls Existence of T-families

Question Given ε > 0 and m ∈ N, does there exist n := n(ε, m) such that whenever (Ω, F, µ) is a probability space and (A{i,j}){i,j}∈[n]2 are such that µ(A{i,j}) > ε for all {i, j} ∈ [n]2, then there is s ∈ [n]m such that

{i,j}∈[s]2 A{i,j} = ∅?

NO: Example by Erd¨

  • s and Hajnal: Fix arbitrary n, r > 0. For

each i, j ∈ [n]2, let A{i,j} := {(ak)k<n ∈ r n : ai = aj}. Consider r n with its counting probability measure, µ(s) = |s|/r n for s ⊆ r n. Then

1 µ(A{i,j}) ≥ 1 − 1/r, and 2 {i,j}∈[s]2 A{i,j} = ∅ for every s ⊆ n with |s| ≥ r + 1.

slide-52
SLIDE 52

On the Banach-Saks property and convex hulls Existence of T-families

Question Given ε > 0 and m ∈ N, does there exist n := n(ε, m) such that whenever (Ω, F, µ) is a probability space and (A{i,j}){i,j}∈[n]2 are such that µ(A{i,j}) > ε for all {i, j} ∈ [n]2, then there is s ∈ [n]m such that

{i,j}∈[s]2 A{i,j} = ∅?

NO: Example by Erd¨

  • s and Hajnal: Fix arbitrary n, r > 0. For

each i, j ∈ [n]2, let A{i,j} := {(ak)k<n ∈ r n : ai = aj}. Consider r n with its counting probability measure, µ(s) = |s|/r n for s ⊆ r n. Then

1 µ(A{i,j}) ≥ 1 − 1/r, and 2 {i,j}∈[s]2 A{i,j} = ∅ for every s ⊆ n with |s| ≥ r + 1.

slide-53
SLIDE 53

On the Banach-Saks property and convex hulls Existence of T-families

Question Given ε > 0 and m ∈ N, does there exist n := n(ε, m) such that whenever (Ω, F, µ) is a probability space and (A{i,j}){i,j}∈[n]2 are such that µ(A{i,j}) > ε for all {i, j} ∈ [n]2, then there is s ∈ [n]m such that

{i,j}∈[s]2 A{i,j} = ∅?

NO: Example by Erd¨

  • s and Hajnal: Fix arbitrary n, r > 0. For

each i, j ∈ [n]2, let A{i,j} := {(ak)k<n ∈ r n : ai = aj}. Consider r n with its counting probability measure, µ(s) = |s|/r n for s ⊆ r n. Then

1 µ(A{i,j}) ≥ 1 − 1/r, and 2 {i,j}∈[s]2 A{i,j} = ∅ for every s ⊆ n with |s| ≥ r + 1.

slide-54
SLIDE 54

On the Banach-Saks property and convex hulls Existence of T-families

Question Given ε > 0 and m ∈ N, does there exist n := n(ε, m) such that whenever (Ω, F, µ) is a probability space and (A{i,j}){i,j}∈[n]2 are such that µ(A{i,j}) > ε for all {i, j} ∈ [n]2, then there is s ∈ [n]m such that

{i,j}∈[s]2 A{i,j} = ∅?

NO: Example by Erd¨

  • s and Hajnal: Fix arbitrary n, r > 0. For

each i, j ∈ [n]2, let A{i,j} := {(ak)k<n ∈ r n : ai = aj}. Consider r n with its counting probability measure, µ(s) = |s|/r n for s ⊆ r n. Then

1 µ(A{i,j}) ≥ 1 − 1/r, and 2 {i,j}∈[s]2 A{i,j} = ∅ for every s ⊆ n with |s| ≥ r + 1.