Chapter 5

Set Systems (Hypergraphs)

[37/70%]
Show
LaTeX
Problem 5.1.1

Prove that any family of pairwise intersecting subsets of an nn-element set contains at most 2n−12^{n-1} subsets.

?
Problem 5.1.2

Let 2⩽t⩽n−22 \leqslant t \leqslant n-2.

?
(1)

Construct a family of 2n−t2^{n-t} subsets of an nn-element set any two of which intersect in at least tt elements.

(2)

Does there exist such a family of 2n−t+12^{n-t}+1 subsets?

Problem 5.1.3

Let F\mathscr {F} be any family of kk-element subsets of an nn-element set.

?
(1)

If 2k⩽n2k \leqslant n and any two subsets from F\mathscr {F} intersect, then ∣F∣⩽(n−1k−1)\left|\mathscr {F}\right| \leqslant \binom {n-1}{k-1}.

(2)

If 2k⩾n2k \geqslant n and the union of no two subsets from F\mathscr {F} is the whole nn-element set, then ∣F∣⩽(n−1k)\left|\mathscr {F}\right| \leqslant \binom {n-1}{k}.

Problem 5.1.4

Any family of twenty 5-element subsets of a 15-element set can be partitioned into 6 subfamilies so that any two disjoint subsets lie in different subfamilies.

?
Remark.
?

In such situations (see Sec. 7.1) one usually speaks of a coloring in different colors instead of a partition into subfamilies. Then, contrary to the intuitive idea of coloring, it is the sets that are colored, while their elements are not colored.

Problem 5.1.5

For l<kl < k, denote by M(n,k,l)M(n,k,l) the minimum number of kk-element subsets of the set Rn\mathscr {R}_{n} such that every ll-element subset of the set Rn\mathscr {R}_{n} is entirely contained in at least one of them. For example, Problem 1.6.2 (1) states that M(n,k,l)⩾(nl)/(kl)M(n,k,l) \geqslant \binom {n}{l} / \binom {k}{l}.

?
(1)

Find M(n,k,1)M(n,k,1).

(2)

Find M(6k+3,3,2)M(6k+3,3,2).

(3)

Find M(n,3,2)M(n,3,2).

(4)

Prove that M(n,k,l)⩾nM(n−1,k−1,l−1)/kM(n,k,l) \geqslant nM(n-1,k-1,l-1)/k.

Problem 5.2.1

A group of students consists of 20 people. Exactly 5 of them are specialists in internet search, 5 in fighting spam, etc., 18 kinds of problems in total (so, obviously, some students are specialists in several problems). It is required to form a strong team of developers from these students. One wants the team to contain a specialist in each problem and the size of the team to be as small as possible (to save on salaries).

?
(1)

Under any arrangement, it is possible to recruit such a team of seven people.

(2)

Under some arrangement, it is not possible to recruit such a team of five people.

Problem 5.2.2

For the collection {{1,6},{1,2},{2,3},{3,4},{4,5},{5,6}}\left\{ \left\{ 1,6\right\} ,\left\{ 1,2\right\} ,\left\{ 2,3\right\} ,\left\{ 3,4\right\} ,\left\{ 4,5\right\} ,\left\{ 5,6\right\} \right\} of sets, find (1) some SCR; (2) an SCR of the smallest size.

?
Remark.
?

A system of common representatives (SCR for short) for a collection M\mathscr {M} of sets is any set AA such that M∩A≠∅M \cap A \neq \emptyset for every M∈MM \in \mathscr {M}.

Problem 5.2.3
?
(1)

Find the smallest size of an SCR for the collection of all kk-element subsets of the set Rn\mathscr {R}_{n}.

(2)

How many SCRs of the smallest size does it have?

Remark.
?

A minimal SCR is an SCR of the smallest size for a given collection M\mathscr {M}. We call an element of ((Rnk)s)\binom {\binom {\mathscr {R}_{n}}{k}}{s} an (n,s,k)(n,s,k)-collection, i.e., a collection of kk-element subsets of the set Rn\mathscr {R}_{n} that contains ss sets. (This term is not generally accepted.)

Problem 5.2.4
?
(1)

Construct a (2n,2(n−1k−1),k)\left(2n, 2\binom {n-1}{k-1}, k\right)-collection for which the minimal SCR consists of two elements and is unique.

(2)

For given n,kn,k, find the largest ss for which there exists an (n,s,k)(n,s,k)-collection having exactly two minimal SCRs.

Problem 5.2.5

The greedy algorithm is the following one. Take any element lying in the maximum number of sets of the given collection. Add it to the ``pre-SCR'' and discard the sets that contain it. Similarly, take an element lying in the maximum number of the remaining sets, etc.

Construct an example of a collection of sets for which the size of a minimal SCR is kk less than the size of any of the SCRs that can be obtained by the greedy algorithm, if (1) k=1k=1; (2) k=2k=2; (3) kk is arbitrary.

?
Problem 5.2.6
?
(1)

For any (n,s,k)(n,s,k)-collection there exists an SCR of size less than

G(n,s,k):=max⁡{nk,nkln⁡skn}+nk+1. G(n,s,k) := \max \left\{ \frac{n}{k}, \frac{n}{k}\ln \frac{sk}{n}\right\} + \frac{n}{k} + 1.
(2)

If n⩾32kn \geqslant 32k and 60⩽skn<ek60 \leqslant \dfrac {sk}{n} < e^{k}, then there exists an (n,s,k)(n,s,k)-collection every SCR of which has size greater than n64kln⁡skn\dfrac {n}{64k}\ln \dfrac {sk}{n}.

(3)

If

k⩽n−landG((nk),(nl),(n−lk))⩽s, k \leqslant n-l \quad \text{and} \quad G\left(\binom {n}{k}, \binom {n}{l}, \binom {n-l}{k}\right) \leqslant s,

then there exists an (n,s,k)(n,s,k)-collection every SCR of which has size greater than ll.

(4)

For all sufficiently large nn, if k2l+kl2<n1,8k^{2}l+kl^{2} < n^{1{,}8}, then k<n−lk < n-l and (nk)/(n−lk)<2ekl/n\binom {n}{k}\Big/\binom {n-l}{k} < 2e^{kl/n}.

(5)

For all sufficiently large nn and kk, if 101ln⁡ln⁡k<ln⁡skn<k<n4101\ln \ln k < \ln \dfrac {sk}{n} < \sqrt{k} < \sqrt[4]{n}, then there exists an (n,s,k)(n,s,k)-collection every SCR of which has size greater than 0,99nkln⁡skn0{,}99\dfrac {n}{k}\ln \dfrac {sk}{n}.

(6)

If

l⩽n−kand(nl)((nk)−(n−lk))<((nk)s), l \leqslant n-k \quad \text{and} \quad \binom {n}{l}\left(\binom {n}{k}-\binom {n-l}{k}\right) < \binom {\binom {n}{k}}{s},

then there exists an (n,s,k)(n,s,k)-collection every SCR of which has size greater than ll.

Problem 5.3.1
?
(1)

Matching lemma. Suppose there are several (finitely many) boys and girls. Each boy loves some (quite possibly zero) number of girls. Then all the boys can be married to girls they love (so that the married couples do not overlap) if and only if for any set of boys, the number of girls loved by at least one of them is not less than the number of these boys.

(2)

Hall’s theorem. Let S1,…,SmS_{1}, \ldots , S_{m} be finite sets. One can choose an element xi∈Six_{i} \in S_{i} in each of them so that all xix_{i} are distinct if and only if for each k∈{1,…,m}k \in \left\{ 1,\ldots ,m\right\} the union of any kk of these sets has at least kk elements.

Problem 5.3.2

What is the minimum number of edges that must be deleted from the graph Kn,nK_{n,n} so that no perfect matchings remain (i.e., subgraphs consisting of nn disjoint segments)?

?
Problem 5.3.3

Suppose that in a system of mm-element sets every element belonging to at least one of them belongs to exactly ll of them. Then for m⩾lm \geqslant l this system of sets has an SDR.

?
Remark.
?

Let a family M\mathscr {M} of sets be given. In each of the sets an element is chosen. If all these elements are distinct, then such a collection is called a system of distinct representatives (SDR for short). Or, formally, a system of distinct representatives for a family M\mathscr {M} of sets is an ordered collection of distinct elements x(S)∈Sx(S) \in S, S∈MS \in \mathscr {M}, i.e., an injective map x:M→⋃S∈MSx : \mathscr {M} \to \bigcup_{S \in \mathscr {M}} S such that x(S)∈Sx(S) \in S for every S∈MS \in \mathscr {M}.

(The ordering of the collection matters for counting the number of SDRs in Problem 5.3.5, not for determining whether an SDR exists.)

For example, Hall's theorem states that a system S1,…,SmS_{1}, \ldots , S_{m} of finite sets has a system of distinct representatives if and only if ∣⋃i∈ISi∣⩾∣I∣\left|\bigcup_{i \in I} S_{i}\right| \geqslant \left|I\right| for every I⊂{1,…,m}I \subset \left\{ 1,\ldots ,m\right\}.

Problem 5.3.4

An SDR of a subfamily can be extended to an SDR of the whole family. Here is a more detailed formulation. From a family M\mathscr {M} of sets, several subsets S1,…,SkS_{1}, \ldots , S_{k} are chosen. Suppose that the elements x1,…,xkx_{1}, \ldots , x_{k} form an SDR of the family of sets S1,…,Sk∈MS_{1}, \ldots , S_{k} \in \mathscr {M}. If the whole family M\mathscr {M} has an SDR, then it has an SDR containing the elements x1,…,xkx_{1}, \ldots , x_{k}.

?
Problem 5.3.5

Denote by F(S1,…,Sm)F(S_{1}, \ldots , S_{m}) the number of SDRs of the system {S1,…,Sm}\left\{ S_{1}, \ldots , S_{m}\right\}.

?
(1)

Is it true that for every kk there exists a system S1,…,SmS_{1}, \ldots , S_{m} such that F(S1,…,Sm)=kF(S_{1}, \ldots , S_{m}) = k?

(2)

Find all possible values of F(S1,S2)F(S_{1}, S_{2}) given that ∣S1∣=∣S2∣=5\left|S_{1}\right| = \left|S_{2}\right| = 5.

(3)

Find all possible values of F(S1,S2,S3)F(S_{1}, S_{2}, S_{3}) given that ∣S1∣=∣S2∣=∣S3∣=5\left|S_{1}\right| = \left|S_{2}\right| = \left|S_{3}\right| = 5.

Problem 5.3.6

Suppose we are given two partitions of a set SS into mm subsets: S=⨆i=1mAi=⨆i=1mBiS = \bigsqcup_{i=1}^{m} A_{i} = \bigsqcup_{i=1}^{m} B_{i}, m⩽∣S∣m \leqslant \left|S\right|. Suppose one of the following conditions holds.

?
(1)

For every subset {i1,…,ik}⊂{1,…,m}\left\{ i_{1},\ldots ,i_{k}\right\} \subset \left\{ 1,\ldots ,m\right\}, the set Ai1∪…∪AikA_{i_{1}} \cup \ldots \cup A_{i_{k}} contains at most kk of the sets B1,…,BmB_{1}, \ldots , B_{m}.

(2)

∣A1∣=…=∣Am∣=∣B1∣=…=∣Bm∣\left|A_{1}\right| = \ldots = \left|A_{m}\right| = \left|B_{1}\right| = \ldots = \left|B_{m}\right|.

Then the sets A1,…,AmA_{1}, \ldots , A_{m} can be renumbered so that in the new numbering Ai∩Bi≠∅A_{i} \cap B_{i} \neq \emptyset for every i=1,…,mi = 1, \ldots , m.

Problem 5.3.7

Suppose we are given two partitions of a set SS into mm subsets: S=⨆i=1mAi=⨆i=1mBiS = \bigsqcup_{i=1}^{m} A_{i} = \bigsqcup_{i=1}^{m} B_{i}, m⩽∣S∣m \leqslant \left|S\right|. Suppose that for any subsets I,J⊂{1,…,m}I, J \subset \left\{ 1,\ldots ,m\right\} the following inequality holds:

∣(⨆i∈IAi)∩(⨆j∈JBj)∣⩾∣I∣+∣J∣−m. \left|\left(\bigsqcup _{i \in I} A_{i}\right) \cap \left(\bigsqcup _{j \in J} B_{j}\right)\right| \geqslant \left|I\right| + \left|J\right| - m.

Then the sets A1,…,AmA_{1}, \ldots , A_{m} can be renumbered so that after the renumbering there exist pairwise distinct elements xi∈Ai∩Bix_{i} \in A_{i} \cap B_{i}, i=1,…,mi = 1, \ldots , m.

?
Remark.
?

Such a collection x1,…,xmx_{1}, \ldots , x_{m} is called a common system of distinct representatives of the families of sets A1,…,AmA_{1}, \ldots , A_{m} and B1,…,BmB_{1}, \ldots , B_{m}.

Problem 5.3.8
?
(1)

Find a necessary and sufficient condition on a bipartite graph under which its vertices can be labeled A1,…,AnA_{1}, \ldots , A_{n} (in the first part) and B1,C1,…,Bn,CnB_{1}, C_{1}, \ldots , B_{n}, C_{n} (in the second part) so that there are edges A1B1,A1C1,…,AnBn,AnCnA_{1}B_{1}, A_{1}C_{1}, \ldots , A_{n}B_{n}, A_{n}C_{n}.

(2)

Suppose there are mm boys and several girls, each boy loves at least tt girls, and all the boys can be married to girls they love (so that the married couples do not overlap), i.e., there is a matching. Then there are at least

{t!,t⩽m,t!/(t−m)!,t>m, \begin{cases} t!, & t \leqslant m, \\ t!/(t-m)!, & t > m, \end{cases}

ways to marry off the boys to girls they love.

Problem 5.4.1

Find the permanent of the matrix

?
(1)

(abcd)\begin{pmatrix} a & b \\ c & d \end{pmatrix};

(2)

4×44 \times 4 in which k=0,1,2,3,4k = 0,1,2,3,4 diagonal entries are zeros and all the other (including the off-diagonal) entries are ones;

(3)

n×nn \times n with zeros on the diagonal and ones off the diagonal.

Remark.
?

The permanent of a square matrix A=(aij)A = (a_{ij}) of size n×nn \times n is defined by the formula

Per⁡(A):=∑σ∈Σn∏i=1nai,σ(i), \operatorname {Per}(A) := \sum _{\sigma \in \Sigma _{n}} \prod _{i=1}^{n} a_{i,\sigma (i)},

where Σn\Sigma_{n} is the set of all permutations of an nn-element set.

Problem 5.4.2

Find the permanent of the m×nm \times n matrix consisting entirely of ones.

?
Remark.
?

A submatrix of a given matrix is a matrix obtained from it by deleting some number of rows and columns. The permanent of a rectangular matrix AA is defined as the sum of the permanents of all its square submatrices of maximal size. Or, as a formula, for m<nm < n

Per⁡(A):=∑σ∏i=1mai,σ(i), \operatorname {Per}(A) := \sum _{\sigma } \prod _{i=1}^{m} a_{i,\sigma (i)},

where the sum is taken over all mm-element arrangements without repetition of the numbers from 1 to nn. For m>nm > n we set Per⁡(A):=Per⁡(AT)\operatorname {Per}(A) := \operatorname {Per}(A^{T}).

Problem 5.4.3
?
(1)

The permanent does not change under a permutation of the rows.

(2)

Row expansion formula. If m⩽nm \leqslant n, then for every ii

Per⁡(A)=∑j=1naijPer⁡(Aij), \operatorname {Per}(A) = \sum _{j=1}^{n} a_{ij}\operatorname {Per}(A_{ij}),

where AijA_{ij} is the matrix obtained from the original one by deleting the ii-th row and the jj-th column.

Problem 5.4.4
?
(1)

The permanent of an n×nn \times n matrix of zeros and ones equals zero if and only if it has a zero s×ts \times t submatrix with s+t=n+1s+t = n+1.

(2)

For any m⩽nm \leqslant n, the permanent of a rectangular m×nm \times n matrix of zeros and ones equals the number of SDRs (Sec. 5.3) of the system of mm subsets of the set {1,…,n}\left\{ 1,\ldots ,n\right\} determined by the rows of this matrix.

Problem 5.5.1
?
(1)

Mathematicians Vasya and Charlie play a game. First, Charlie marks kk points in the plane. Then Vasya colors some of these points. If Charlie can now draw a line separating the colored points from the uncolored ones, he wins; otherwise he loses. What is the largest kk for which Charlie can win regardless of Vasya's moves?

(2)

The same, but the points are marked in space, and Charlie draws a plane.

Remark.
?

Let R⊂2X\mathscr {R} \subset 2^{X} be a family of subsets of an arbitrary set XX. The Vapnik--Chervonenkis dimension VC⁡(X,R)\operatorname {VC}(X,\mathscr {R}) (or VC-dimension) of the pair (X,R)(X,\mathscr {R}) is the maximal nn such that there exists an nn-element subset A⊂XA \subset X for which every subset of AA is the intersection of AA with some set from R\mathscr {R}. Such a subset AA is said to be shattered by the system R\mathscr {R}. If no such nn exists, one sets VC⁡(X,R):=∞\operatorname {VC}(X,\mathscr {R}) := \infty.

Problem 5.5.2
?
(1)

Find the VC-dimension of the family of all (two-dimensional closed) rectangles in the plane with sides parallel to the coordinate axes.

(2)

Theorem. The VC-dimension of the family of all half-spaces in Rn\mathbb {R}^{n} equals n+1n+1.

(3)

Radon’s theorem. Any n+2n+2 points in Rn\mathbb {R}^{n} can be partitioned into two sets whose convex hulls intersect.

Problem 5.5.3

Find the VC-dimension of the following families of sets:

?
(1)

{{1,…,k}:k∈N}\left\{ \left\{ 1,\ldots ,k\right\} : k \in \mathbb {N}\right\};

(2)

{{k,k+1,k+2,…}:k∈N}\left\{ \left\{ k,k+1,k+2,\ldots \right\} : k \in \mathbb {N}\right\};

(3)

{{k,2k,3k,…}:k∈N}\left\{ \left\{ k,2k,3k,\ldots \right\} : k \in \mathbb {N}\right\};

(4)

{{k1k2,2k1k2,3k1k2,…}:k1,k2∈N}\left\{ \left\{ k_{1}k_{2}, 2k_{1}k_{2}, 3k_{1}k_{2},\ldots \right\} : k_{1},k_{2} \in \mathbb {N}\right\};

(5)

{{k,k2,k3,…}:k∈N}\left\{ \left\{ k,k^{2},k^{3},\ldots \right\} : k \in \mathbb {N}\right\}.

Problem 5.5.4

Find the VC-dimension of the following finite families:

?
(1)

{1,2,3}\left\{ 1,2,3\right\}, {4,5,6}\left\{ 4,5,6\right\}, {1,2,4}\left\{ 1,2,4\right\}, {1,2,5}\left\{ 1,2,5\right\}, {2,3,6}\left\{ 2,3,6\right\}, {3,4,5}\left\{ 3,4,5\right\}, {3,4,6}\left\{ 3,4,6\right\}, {2,4,5,6}\left\{ 2,4,5,6\right\};

(2)

{1,2,3}\left\{ 1,2,3\right\}, {3,4,5}\left\{ 3,4,5\right\}, {1,2,4}\left\{ 1,2,4\right\}, {1,2,5}\left\{ 1,2,5\right\}, {2,4,5}\left\{ 2,4,5\right\}, {2,3,5}\left\{ 2,3,5\right\}, {2,6,7}\left\{ 2,6,7\right\}, {3,4,6,7}\left\{ 3,4,6,7\right\};

(3)

{1,2,3}\left\{ 1,2,3\right\}, {4,5,6}\left\{ 4,5,6\right\}, {7,8,9}\left\{ 7,8,9\right\}, {1,4,7}\left\{ 1,4,7\right\}, {2,5,8}\left\{ 2,5,8\right\}, {3,6,9}\left\{ 3,6,9\right\}, {1,6,9}\left\{ 1,6,9\right\};

(4)

Is it possible to add one more set to the systems from the previous parts so that the VC-dimension increases by 1?

Problem 5.5.5
?
(1)

Is the equality VC(R2,R)=∞VC(\mathbb {R}^{2}, \mathscr {R}) = \infty possible for some family R⊂2R2\mathscr {R} \subset 2^{\mathbb {R}^{2}}?

(2)

The same for some countable family R⊂2R2\mathscr {R} \subset 2^{\mathbb {R}^{2}} of bounded sets.

Problem 5.5.6

In every family of VC-dimension dd each of whose sets has at most rr elements, there exist subsets XX and YY such that

?
(1)

∣X∩Y∣⩽r−d\left|X \cap Y\right| \leqslant r-d;

(2)

∣X∩Y∣⩾d−1\left|X \cap Y\right| \geqslant d-1.

Problem 5.5.7

If R⊂2Rn\mathscr {R} \subset 2^{\mathscr {R}_{n}} and ∣R∣=n\left|\mathscr {R}\right| = n, then for every k=1,2,…,nk = 1,2,\ldots ,n there exists a set AA such that ∣A∣=k−1\left|A\right| = k-1 and ∣{R∩A:R∈R}∣⩾k\left|\left\{ R \cap A : R \in \mathscr {R}\right\} \right| \geqslant k.

?
Problem 5.5.8
?
(1)

If R⊂2Rn\mathscr {R} \subset 2^{\mathscr {R}_{n}} is a family of VC-dimension dd, then there exists a hereditary (i.e., containing, together with each set, all of its subsets) family R′⊂2Rn\mathscr {R}' \subset 2^{\mathscr {R}_{n}} of VC-dimension dd for which ∣R′∣⩽∣R∣\left|\mathscr {R}'\right| \leqslant \left|\mathscr {R}\right|.

(2)

The same, but with ∣R′∣⩾∣R∣\left|\mathscr {R}'\right| \geqslant \left|\mathscr {R}\right|.

Problem 5.5.9

If R⊂2Rn\mathscr {R} \subset 2^{\mathscr {R}_{n}}, then

∣R∣⩽(n0)+(n1)+(n2)+…+(nVC(Rn,R)). \left|\mathscr {R}\right| \leqslant \binom {n}{0} + \binom {n}{1} + \binom {n}{2} + \ldots + \binom {n}{VC(\mathscr {R}_{n},\mathscr {R})}.
?
Problem 5.6.1

Find the size of a minimal SCR (Sec. 5.2) of a sunflower.

?
Remark.
?

A sunflower with kk petals and core YY is a collection of sets {F1,…,Fk}\left\{ F_{1},\ldots ,F_{k}\right\} such that ∣Fi∩Fj∣=Y\left|F_{i} \cap F_{j}\right| = Y for i≠ji \neq j and all the sets Fi∖YF_{i} \setminus Y are nonempty. For example,

  • pairwise disjoint sets form a sunflower with an empty core;

  • one-dimensional vector subspaces (Sec. 7.1) form a sunflower with a one-element core;

  • the sets XpX_{p} of all rational fractions with a fixed prime denominator pp form a sunflower with an infinite core.

Most of the problems in this section are taken from the book [J].

Problem 5.6.2
?
(1)

If ∣F∣>s!(k−1)s\left|\mathscr {F}\right| > s!(k-1)^{s}, then F\mathscr {F} contains a sunflower with kk petals.

(2)

There exist (k−1)s(k-1)^{s} subsets of a finite set, each of which has ss elements and among which one cannot choose a sunflower with kk petals.

Problem 5.6.3

For every prime pp there exists a weak Δ\Delta-system of (p+1)(p+1)-element sets that is not a sunflower and consists of p2+p+1p^{2}+p+1 sets.

?
Remark.
?

A weak Δ\Delta-system is a collection of sets S1,…,SkS_{1},\ldots ,S_{k} such that the numbers ∣Si∩Sj∣\left|S_{i} \cap S_{j}\right| are the same for all i≠ji \neq j. Obviously, a sunflower is a weak Δ\Delta-system. The converse is false. However, the following theorem holds.

Deza’s theorem. If F\mathscr {F} is a weak Δ\Delta-system of ss-element sets and ∣F∣⩾s2−s+2\left|\mathscr {F}\right| \geqslant s^{2}-s+2, then F\mathscr {F} is a sunflower.

The proof of the theorem is rather difficult.

Let us show that the above bound is sharp.

Problem 5.6.4
?
(1)

If F\mathscr {F} is a finite collection of ss-element sets and ∣F∣>(k−1)s\left|\mathscr {F}\right| > (k-1)^{s}, then there exist kk sets from F\mathscr {F} whose common part has fewer than ss elements.

(2)

The bound ∣F∣>(k−1)s\left|\mathscr {F}\right| > (k-1)^{s} in part (1) is sharp.

Remark.
?

The common part of sets S1,…,SkS_{1},\ldots ,S_{k} is the union ⋃i≠j(Si∩Sj)\bigcup_{i \neq j}(S_{i} \cap S_{j}) of all their pairwise intersections.

Problem 5.6.5
?
(1)

If F\mathscr {F} is a finite collection of ss-element sets and ∣F∣>(k−1)s\left|\mathscr {F}\right| > (k-1)^{s}, then F\mathscr {F} contains a flower with kk petals (and some core).

(2)

The bound ∣F∣>(k−1)s\left|\mathscr {F}\right| > (k-1)^{s} from part (1) is sharp.

Remark.
?

A flower with kk petals and core YY is a collection F\mathscr {F} of sets such that each of them contains YY and there is no SCR of k−1k-1 elements for the collection {S∖Y:S∈F}\left\{ S \setminus Y : S \in \mathscr {F}\right\}.