The lemma Corradi , English Corradi's lemma , named after the Hungarian mathematician Kereszyély Corradi is a theorem that the transition field of the mathematical part areas combinatorics , Graph Theory and finite geometry is attributable. The lemma provides information about the minimum number of nodes that a finite uniform hypergraph must have, if it is assumed that two different hyper-edges have a certain number of nodes in common .
Formulation of the lemma
In detail, the lemma says:
Be natural numbers and be an integer with .
n
,
r
,
N
{\ displaystyle n, r, N}
k
{\ displaystyle k}
n
≥
r
≥
k
≥
0
{\ displaystyle n \ geq r \ geq k \ geq 0}
Let be a uniform hypergraph, consisting of a set of nodes with nodes and a - membered family of hyperedges with nodes each .
H
=
(
X
,
A.
)
{\ displaystyle H = (X, {\ mathcal {A}})}
X
{\ displaystyle X}
n
{\ displaystyle n}
N
{\ displaystyle N}
A.
=
(
A.
1
,
A.
2
,
...
,
A.
N
)
{\ displaystyle {\ mathcal {A}} = (A_ {1}, A_ {2}, \ ldots, A_ {N})}
r
{\ displaystyle r}
Let it be further assumed that the number condition is always given for all intersections of two hyperedges .
A.
i
∩
A.
j
(
i
,
j
=
1
,
2
,
...
,
N
)
{\ displaystyle A_ {i} \ cap A_ {j} \; (i, j = 1,2, \ ldots, N)}
|
A.
i
∩
A.
j
|
≤
k
{\ displaystyle | A_ {i} \ cap A_ {j} | \ leq k}
Then:
(*) .
n
≥
N
r
2
r
+
(
N
-
1
)
k
{\ displaystyle n \ geq {\ frac {Nr ^ {2}} {r + (N-1) k}}}
proof
The proof of the inequality (*) results - in connection with the representation by Jukna and Lovász - by applying the method of double counting and the other inequality in the following steps:
Determinations
(1)
d
(
x
)
: =
deg
H
(
x
)
=
|
{
i
∈
{
1
,
2
,
...
,
N
}
:
x
∈
A.
i
}
|
(
x
∈
X
)
{\ displaystyle d (x): = {\ deg _ {H}} (x) = | \ {i \ in \ {1,2, \ ldots, N \} \ colon x \ in A_ {i} \} | \; (x \ in X)}
(2)
J
Y
: =
{
(
y
,
i
)
∈
Y
×
{
1
,
2
,
...
,
N
}
:
y
∈
A.
i
}
(
Y
⊆
X
)
{\ displaystyle J_ {Y}: = \ {(y, i) \ in Y \ times \ {1,2, \ ldots, N \} \ colon y \ in A_ {i} \} \; (Y \ subseteq X)}
(3)
K
: =
{
(
x
,
i
,
j
)
∈
X
×
{
1
,
2
,
...
,
N
}
×
{
1
,
2
,
...
,
N
}
:
x
∈
A.
i
∩
A.
j
}
{\ displaystyle K: = \ {(x, i, j) \ in X \ times \ {1,2, \ ldots, N \} \ times \ {1,2, \ ldots, N \} \ colon x \ in A_ {i} \ cap A_ {j} \}}
Double counting
(4)
|
J
Y
|
=
∑
y
∈
Y
d
(
y
)
=
∑
j
=
1
N
|
Y
∩
A.
j
|
(
Y
⊆
X
)
{\ displaystyle | J_ {Y} | = \ sum _ {y \ in Y} {d (y)} = \ sum _ {j = 1} ^ {N} {| Y \ cap A_ {j} |} \ ; (Y \ subseteq X)}
(5)
|
K
|
=
∑
x
∈
X
d
(
x
)
2
=
∑
i
,
j
=
1
N
|
A.
i
∩
A.
j
|
=
∑
i
=
1
N
|
J
A.
i
|
{\ displaystyle | K | = \ sum _ {x \ in X} {d (x) ^ {2}} = \ sum _ {i, j = 1} ^ {N} {| A_ {i} \ cap A_ {j} |} = \ sum _ {i = 1} ^ {N} {| J_ {A_ {i}} |}}
Upward estimate
With (4) results in particular for :
i
=
1
,
2
,
...
,
N
{\ displaystyle i = 1,2, \ ldots, N}
(6)
|
J
A.
i
|
=
∑
j
=
1
N
|
A.
i
∩
A.
j
|
=
|
A.
i
|
+
∑
j
=
1
,
2
,
...
,
N
i
≠
j
|
A.
i
∩
A.
j
|
≤
r
+
(
N
-
1
)
k
{\ displaystyle | J_ {A_ {i}} | = \ sum _ {j = 1} ^ {N} {| A_ {i} \ cap A_ {j} |} = | A_ {i} | + \ sum _ {j = 1,2, \ ldots, N \ atop \; i \ neq j} {| A_ {i} \ cap A_ {j} |} \ leq r + (N-1) k}
So it follows from (5) and (6) :
(7)
∑
x
∈
X
d
(
x
)
2
≤
N
(
r
+
(
N
-
1
)
k
)
{\ displaystyle \ sum _ {x \ in X} {d (x) ^ {2}} \ leq {N {\ bigl (} r + (N-1) k {\ bigr)}}}
Downward estimate
By virtue of those inequality we get:
(8th)
∑
x
∈
X
d
(
x
)
2
=
n
⋅
∑
x
∈
X
1
n
⋅
d
(
x
)
2
≥
n
⋅
(
∑
x
∈
X
1
n
⋅
d
(
x
)
)
2
=
n
⋅
1
n
2
⋅
(
∑
x
∈
X
d
(
x
)
)
2
=
1
n
⋅
(
∑
x
∈
X
d
(
x
)
)
2
{\ displaystyle \ sum _ {x \ in X} {d (x) ^ {2}} = n \ cdot \ sum _ {x \ in X} {{\ frac {1} {n}} \ cdot d ( x) ^ {2}} \ geq n \ cdot {\ bigl (} \ sum _ {x \ in X} {{\ frac {1} {n}} \ cdot d (x)} {\ bigr)} ^ {2} = n \ cdot {\ frac {1} {n ^ {2}}} \ cdot {\ bigl (} \ sum _ {x \ in X} {d (x)} {\ bigr)} ^ { 2} = {\ frac {1} {n}} \ cdot {\ bigl (} \ sum _ {x \ in X} {d (x)} {\ bigr)} ^ {2}}
Again with (4) and then follows:
Y
=
X
{\ displaystyle Y = X}
(9)
∑
x
∈
X
d
(
x
)
2
≥
1
n
⋅
|
J
X
|
2
=
1
n
⋅
(
∑
j
=
1
N
|
A.
j
|
)
2
=
(
N
r
)
2
n
=
N
2
r
2
n
{\ displaystyle \ sum _ {x \ in X} {d (x) ^ {2}} \ geq {\ frac {1} {n}} \ cdot {| J_ {X} |} ^ {2} = { \ frac {1} {n}} \ cdot {\ bigl (} \ sum _ {j = 1} ^ {N} {| A_ {j} |} {\ bigr)} ^ {2} = {\ frac { (No) ^ {2}} {n}} = {\ frac {N ^ {2} r ^ {2}} {n}}}
Summary
(7) and (9) together then result in:
(10)
N
(
r
+
(
N
-
1
)
k
)
≥
N
2
r
2
n
{\ displaystyle {N {\ bigl (} r + (N-1) k {\ bigr)}} \ geq {\ frac {N ^ {2} r ^ {2}} {n}}}
Since (10) and (*) are equivalent, everything is shown.
additive
The above inequality (*) is sharp in the sense that finite structures exist for which in the inequality (*) even the equal sign applies. An example of this is offered by every finite projective plane in which one understands its points as nodes and its straight lines as hyper-edges of a hypergraph.
Note on the history
The lemma goes back to a problem that K. Corrádi posed on the occasion of the 1968 Miklós Schweitzer competition for students .
Sources and background literature
K. Corrádi: Problem at Schweitzer competition . In: Matematikai Lapok . tape 20 , 1969, p. 159-162 .
Stasys Jukna: Extremal Combinatorics (= Texts in Theoretical Computer Science ). 2nd Edition. Springer Verlag , Heidelberg, Dordrecht, London, New York 2011, ISBN 978-3-642-17363-9 . MR2865719
László Lovász: Combinatorial Problems and Exercises (= Texts in Theoretical Computer Science ). North-Holland Publishing Company , Amsterdam, New York, Oxford 1979, ISBN 0-444-85219-0 . MR0537284
Gábor J. Székely (Ed.): Contests in Higher Mathematics . Miklós Schweitzer Competitions 1962-1991 (= Problem Books in Mathematics ). Springer Verlag, New York (et al.) 1996, ISBN 0-387-94588-1 . MR1416130
Individual evidence
↑ a b c Stasys Jukna: Extremal Combinatorics. 2011, pp. 23, 37
↑ a b c László Lovász: Combinatorial Problems and Exercises. 1979, pp. 78,130,446-447
↑ Stasys Jukna: Extremal Combinatorics. 2011, p. 37
^ Gábor J. Székely: Contests in Higher Mathematics: Miklós Schweitzer Competitions 1962-1991. 1996, p. 10
<img src="https://de.wikipedia.org/wiki/Special:CentralAutoLogin/start?type=1x1" alt="" title="" width="1" height="1" style="border: none; position: absolute;">