6

Problem Statement:

We have $n$ distinct positive integers say $a_1,a_2....a_n$ and a given integer value $M$. We have to count number of ways to distribute these integers to $r$ identical bins subject to the constraints that the product of integers in any bin can not exceed $M$ and also no bin should be empty and we must distribute all the integers to some bin or the other.

My thoughts:

I believe this is a type of distribution of distinct objects into identical bins but with a constraint on each bin. The constraint function here is the product of integers in the bin, but it can be sum of integers in a bin or number of integers in a bin or anything else. Like say this or this the problem's limiting function is number of balls in a bin. But unlike our case in both of them balls are identical

Approach1 on top of my mind:

As point out by @Sil If the problem was about limit on sum the following approach would have worked though not very efficient.

Lets say my integers are $2,5,7,11$.

If $r=1$ then its equivalent to counting sum of all coefficients of $x^i$ for $1 \le i\le M$ in $(1+x^2)(1+x^5)(1+x^7)(1+x^{11})$

Similarly if $r=2$ then its equivalent to counting sum of all coefficients of $x_1^ix_2^j$ for $1 \le i,j\le M$ and in $(1+x_1^2+x_2^2)(1+x_1^5+x_2^5)(1+x_1^7+x_2^7)(1+x_1^{11}+x_2^{11})$

......

......

Similarly if $r=r$ then its equivalent to counting sum of all coefficients of $x_1^ix_2^j..x_r^w$ for $1 \le i,j,..w\le M$ and in $(1+x_1^2+x_2^2+..+x_r^2)(1+x_1^5+x_2^5+..+x_r^5)(1+x_1^7+x_2^7+..+x_r^7)(1+x_1^{11}+x_2^{11}+..+x_r^{11})$

This is $O(r^n)$, not very efficient to say the least, so there surely exists more efficient ways to get the result.

Another way of thinking is:

I think it can be solved by PIE let's call the resut $f(r$, set of all bins$)$, we count unrestricted, then fix a particular bins split its restrcted and unrestricted case an then call $f(r-1$, remaining $r-1$ bins$)$ , now fix two boxed and add back the same $f(r-2$, remaining $r-2$ bins$)$ and so on. My idea are still vauge, needs little bit more thought.

Another problem:

I have another part of this problem, let me know if solving this is easier. Lets say we have the same numbers $a_1,a_2....a_n$ and the same $M$. Now if I say we dont care about number of bins you use. Form new set of numbers fusing together(i.e product I mean) two or more of them such that no such fused number is greater than M. How many such sets can you form. Here we must use all the numbers.

I was initially thinking about solving the first part and then sum for each r=1 to r for this part but now I feel it might be easier to approach this directly.

Example clarifying this part: intial_nos = {2,5,7,11}, M =70

case 1:{2,5,7,11}

case 2:{10,7,11}

case 3:{5,14,11}

case 4:{5,7,22}

case 5:{2,11,35}

case 6:{2,7,55}

case 7:{70,11}

So answer should be 7

  • 1
    @Sil my bad it indeed addresses the sum not the product. I will update that part. So essentially now I dont have any approach at all. :( –  Nov 04 '22 at 04:25
  • @Sil I have another part of this problem, let me know if solving that is easier. I have updated question with it at the bottom. –  Nov 04 '22 at 04:36
  • Yeah @Sil are my question turning out to be too difficult for math.stackexchange ? should I move them to mathoverflow ? The reason I am asking is I didnt get any satisfactory answer on my earlier question either . –  Nov 04 '22 at 05:36
  • @Dan tell me which case you have the solution for ? You can edit it to anyone the two you like :D –  Nov 04 '22 at 09:35
  • @MasrukaJannat I don't have the solution to either case. – Dan Nov 04 '22 at 10:17
  • It seems to me the problem is equivalent to the sum problem from the theoretical point of view, simply take the condition $\prod a_i \leq M$ and replace it with equivalent $\sum \ln a_i \leq \ln M$ which works fine since all involved values are positive. Only downside is these values are irrational, so practical use is questionable. – Sil Nov 04 '22 at 13:51
  • @Sil Product of numbers grows very fast compared to sum, as you can see in my example we only managed to find product of three numbers once and nothing beyond. So from the point of implementing the brute force method via computer program taking $logM$ would be useful to know how big M is so that we can know how many products we might get in worst case. –  Nov 04 '22 at 14:43

0 Answers0