Sly Subsets of S

Let S = { 1 , 2 , 3 , 12 } S = \{ 1, 2, 3, \ldots 12\} and T 1 , T 2 , T a T_1, T_2, \ldots T_a be subsets of S S such that T i ⊄ T j i j T_i \not \subset T_j \, \forall i \neq j . What is the maximum possible value of a a ?


The answer is 924.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

10 solutions

Let T = { T 1 , T 2 , , T a } , T i = k i , i = 1 , a T=\{T_1,T_2,\ldots,T_a\}, |T_i|=k_i,i=\overline{1,a} .

Note that there are n ! n! permutations of the elements of S S .

We will say that a permutation π \pi of the elements of S S begins with T i T_i if the first k i k_i members of π \pi are the elements of T i T_i in some order. The number of permutations beginning with T i T_i must be k i ! ( 12 k i ) ! k_i!(12-k_i)! .

Because of T i T j , i j T_i\nsubseteq T_j,\forall i\ne j , no permutation can begin with two different sets in T T ; therefore permutations beginning with different sets in T T are distinct.

Therefore: i = 1 a k i ! ( 12 k i ) ! 12 ! \displaystyle\sum_{i=1}^a k_i!(12-k_i)!\le 12! or i = 1 a 1 ( 12 k i ) 1 \displaystyle\sum_{i=1}^a \dfrac{1}{\displaystyle\binom{12}{k_i}}\le 1 .

We have ( 12 k i ) ( 12 6 ) \displaystyle\binom{12}{k_i}\le \binom{12}{6} , so a ( 12 6 ) = 924 \displaystyle a\le\binom{12}{6}=924

Consider T 1 , T 2 , , T 924 S T_1,T_2,\ldots,T_{924}\subset S such that T i = 6 , i |T_i|=6,\forall i . It's easy to look that this subsets such that T i T j , i j T_i\nsubseteq T_j,\forall i\ne j .

This was the only correct solution submitted.

Most solutions took for granted that the sets must all be the same size, which is not obvious.

The result of this question is a direct consequence of Sperner's Theorem (not to be confused with Sperner's Lemma) and can be easily proved using this theorem.

Calvin Lin Staff - 7 years ago

Many people seem to be assuming that all the sets are of equal cardinality. In fact, we can take an algorithmic approach and design a "trickling process" that keeps the size of the antichain strictly increasing while pushing the sets to the "central size". A nice corollary is that we can characterize all extremal families- and they indeed consist of sets of equal cardinality! :D

The details are found here: https://drive.google.com/file/d/0B_QyOg5Kp1eHRnlfbXJibTI0X1E/view?usp=sharing

One interesting aspect of this proof is that it avoids the LYM inequality, and in a sense is easier to "visualize."

Rafay Ashary - 4 years, 1 month ago
Aman Gupta
Oct 21, 2013

Since 12 C k ^{12}C_k is maximum for k = 6 k=6 , hence dividing 12 12 elements into the sets such that each element contains exactly 6 6 elements, is the best way to maximize a a

One thing that I am not able to prove mathematically is :- Why won't any a > 924 a > 924 would violate the subset condition. Someone Kindly help me out ^_^

Well, so far I can only get that in order not to violate T i T j T_i ⊄ T_j , if we include a set T i T_i of size n n , we cannot include ( 12 n 1 ) 12 \choose n-1 subsets. For instance, if we try to include a larger set in T 6 T_6 , (remember that we want to make it like larger ) then that would be a superset of a subset of T 6 T_6 , violating the subset conditions.

Not quite sure whether it is correct though.

Anqi Li - 7 years, 7 months ago

Log in to reply

The proof is not immediately obvious on how to proceed. In your construction, it's possible for distinct T i T_i to share the same subset, and hence we cannot easily extend it. What we want is to relate T i T_i to something that is uniquely determined.

This is actually a result known as Sperner's Theorem, which is a special case of Dilworth's Theorem, which is a special application of the Pigeonhole Principle.

Calvin Lin Staff - 7 years, 7 months ago

Log in to reply

a result known as Sperner's Theorem

Hmm I proved this theorem using Lubell-Yamamoto-Meshalkin inequality. Refer to LYM inequality .

Anqi Li - 7 years, 7 months ago

Indeed, finding 924 lower bound is very easy (took me half a minute) but proving that there is no better solution seems to be hard. I didn't manage to do it yet after trying 5 different approaches based on induction, representing the subsets as binary numbers and much more. I think the most promising approach is somehow to show that any collection having n n subsets eliminates at least n n 6-element subsets (I call a set eliminated if it either contains or is contained in some subset of the collection). This seems believable and if correct, it proves that 924 is also an upper bound.

Anyway, the combination of the two facts -- very easy to guess the solution but hard to prove it -- make this a very poor problem for brilliant in my opinion...

Marek Bernat - 7 years, 7 months ago

Log in to reply

This is a poor problem for use as a pure integer answer question, but your activity on Brilliant is so much more than "Can I guess the answer?". Many other problems are also 'easily guessed in 3 tries', but that doesn't necessarily mean that they should not be used as problems. We do our best to work within the limitations of the system.

This question does illustrate a well known theorem, that can be difficult to approach if you do not know what to do. We felt that the Brilliant community should be exposed to this beautiful piece of mathematics.

Calvin Lin Staff - 7 years, 7 months ago

Log in to reply

Thanks for your explanation. I agree with you that it's worthwhile to introduce people to stuff like this (I looked up the references you mentioned in the other comment -> amazing) but here the main problem is that nobody really bothered to write down a proper solution (or indeed, most people don't even know it, since they just guessed the answer).

That means that until you came around and mentioned those references, I didn't learn from this problem anything at all. Could there be a way to remedy this? E.g. posting an official (i.e. by you) solution for interested readers? Or, if that's too much work, at least those references to set people on the right track.

Marek Bernat - 7 years, 7 months ago
Sai Avinash Kommi
May 20, 2014

Given, S={1,2,3,…12} and T1,T2,…Ta are its subsets S such that Ti⊄Tj for all i≠j. so we must either chose subsets from S such that no subset is a subset of the other if we choose all subsets with size 1 we cant chose subsets with size 2. if we choose all subsets with size 1 we cant chose subsets with size 3. and so on Number of subsets of size r from a set of size n is nCr we need to find to find the maximum value of nCr if n=12 max value of 12Cr is when r=12/2=6 number of subsets = 12C6=924

No valid reason all must be the same size.

Calvin Lin Staff - 7 years ago
Anju Lawrence
May 20, 2014

The total number of subsets that can be formed out of the superset S S which meets the below the criteria
\rightarrow T i ⊄ T j i j T_i \not \subset T_j \, \forall i \neq j

can be denoted as

( N R ) {N \choose R}

where R R indicates cardinality of each subset and R R can take values from 1 t o N {1 to N} . N N denotes the cardinality of the superset S S .

In the given problem where N = 12 N=12 the maximum possible value of ( 12 R ) {12 \choose R} needs to be evaluated. This turns out to be when R = 6 R=6 . Hence the total number of subsets is ( 12 6 ) = 924 {12 \choose 6} = 924

No reason all must be the same size.

Calvin Lin Staff - 7 years ago

We want to have the greatest number of subsets of S that are not subsets of other subsets of S. To accomplish this, we must have the most number of elements in each set to ensure diversity. The best way to do this is to have subsets with the same number of elements such that all of these combinations are unique. Since there are twelve elements, the greatest number of combinations is 12C6 or 12C7 (same value). These will then define the greatest possible number of subsets. Solving, we get 924.

No valid reason all must be the same size.

Calvin Lin Staff - 7 years ago
黎 李
May 20, 2014

C(12,6)=924

Not a proof.

Calvin Lin Staff - 7 years ago

(sorry for my english mistakes) There is a very famous theorem known as "sperner theorem" that I'll prove in the next lines : Let A and B be two "weird" sets if A is not a subset of B nor B is a subset of A. Let X be a family of subsets of {1,2,...,n} such that these subsets are all weird one with other. Then : \sum_{A is a element of X}^{a} {n \choose |A|}^{-1} \le 1

Incomplete solution. It just ends abruptly.

Calvin Lin Staff - 7 years ago
Sam Song
May 20, 2014

We want the maximum number of disjoint subsets of S. If the number of elements in T i T_i is less than the number of elements in T j T_j , then T i T_i can potentially be a subset of T j T_j . Therefore we want all T i T_i 's to have the cardinality (number of elements). This is maximized when we select subsets with cardinality 6 6 . This is because ( 12 6 ) > ( 12 n ) {12 \choose 6} > {12 \choose n} , where n 6 n \neq 6 . Therefore our answer is ( 12 6 ) = 924 {12 \choose 6} = 924 .

The sets aren't disjoint. Having sets of different sizes means one could be contained in another, but that doesn't mean we should choose all the same size.

Calvin Lin Staff - 7 years ago
Chin Siang Ser
May 20, 2014

A typical combination problem. From the pascal triangle,the 12Cr with the biggest value is when r is in the middle of 0-12.Hence, 12C6 is the answer.

No reason all must be the same size.

Calvin Lin Staff - 7 years ago
Zichu Ye
May 20, 2014

We can consider a set of primes P = { p 1 , p 2 , p 3 , . . . , p 12 } P = \{p_1, p_2, p_3, ..., p_{12}\} . We will find as many products (of elements of P P ) that are mutually coprime as possible. It should be intuitive that the products consist of the same number of primes, for instance taking all possible products of three primes we will have p 1 p 2 p 3 , p 2 p 3 p 4 , p 3 p 4 p 5 p_1 p_2 p_3, p_2 p_3 p_4, p_3 p_4 p_5 , etc. Now we should not find a product of more than 3 primes since two or more of the products of 3 primes will always be factors, and products of less than 3 primes will also be factors of two or more of the products of 3 primes. Introducing products of different number of primes will at least sacrifice one product from the count. Then, we simply look for all the possible ways to find a certain number n n of primes from P P to multiply together, that is ( 12 n ) = 12 ! n ! ( 12 n ) ! {12 \choose n} = \frac{12!}{n!(12-n)!} . It has a maximum when n = 6 , ( 12 n ) = 924 n = 6, {12 \choose n} = 924 . The original problem should behave similarly, when we consider the factor of 1 1 to prime numbers as the subset of \varnothing to sets.

No reason the products should have the same number of primes as factors.

Calvin Lin Staff - 7 years ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...