The Student Room Group

The Proof is Trivial!

Scroll to see replies

Original post by jjpneed1
That's pretty harsh Khallil I'm sure you've made those kinds of mistakes at some point :tongue: To be fair your problem isn't very interesting, in fact if the beta function was taught at A-level I'm sure I'd see it in exercise 1A

Fun one but fairly simple:

Problem 474 *

I can write any (positive) number as a sum of arbitrarily many (positive) real numbers, e.g. 16 = 10 + 6 = 3.4 + 6.6 + 6, etc. When is the product of all elements in this sum maximised?


A partial solution, if any 2 of the numbers are different, call them y-e and y+e, then (y-e)(y+e)=y^2-e^2 so making them equal will only increase the product, keeping the sum the same. Repeating this we see that we can continue to increase the product indefinately, and the terms tend to the same number. This isn't rigorous but it seems obvious (from the smoothness of the process) that having them all the same is the way to maximise the product. So if x is our number, we need to maximise (x/N)^N where N is an integer.
Original post by Smaug123
Hmm. I can Lagrange-multipliers it down to "Maximise aia^i subject to a>0, i a natural number, ai=16a i = 16", anyway, because Lagrange multipliers on abca b c subject to a+b+c=16a+b+c=16 yields a=b=ca=b=c. Then the only problem is to find how many terms are in the sum.

That is, maximise (16i)i(\frac{16}{i})^i for integer i. Let's move to the more general xx as our sum (so x=16 above). Now, if i>x then we start raising something less than 1 to a high power, so that can't be a maximum. Our answer must therefore be less than x. Using i=1 gives us x straight off.

Suppose we'd found i. What would we know? We'd have that xii>xi+1i+1\frac{x}{i}^i > \frac{x}{i+1}^{i+1}, or 1ii>1i+1i+1x\frac{1}{i}^i > \frac{1}{i+1}^{i+1}x, or that x<i+1ii(i+1)x < \frac{i+1}{i}^i (i+1). That fraction term tends to ee from below as i increases; in particular, we need i+1>xei+1 > \frac{x}{e}, or i>xe1i > \frac{x}{e}-1. That suggests using i = roundup(x/e - 1).

How tight is this bound? Not sure. It is conceivable that the "fraction term tends to…" line makes the bound slacker than it needs to be.


A quick check with a computer suggests that your value is correct. For x=1000 I get the max occurs at 368 terms. I think this can be made rigorous by differentiating and noting that the function is nice enough that no other itnegers will be better.

EDIT: Differentiation does work, and it is indeed what you said.
(edited 9 years ago)
James and Smaug, you have the main parts even if it's a bit hand-wavey.

Spoiler

(edited 9 years ago)
Original post by jjpneed1
James and Smaug, you have the main parts even if it's a bit hand-wavey.

Spoiler



AM-GM does work better.

Calculus only gives the max for continuous n. To actually prove that the max is when n is one of floor(x/e) or cel(x/e) is a bit harder. You need to show that the function is decreasing either side of the maximum (not difficult, but not trivial either). I see no reason why the answer must be the same choice of floor or cel, I think it will depend of x as the maximum could be closer to one than the other for same values, but the other way round for others.
Original post by james22
AM-GM does work better.

Calculus only gives the max for continuous n. To actually prove that the max is when n is one of floor(x/e) or cel(x/e) is a bit harder. You need to show that the function is decreasing either side of the maximum (not difficult, but not trivial either). I see no reason why the answer must be the same choice of floor or cel, I think it will depend of x as the maximum could be closer to one than the other for same values, but the other way round for others.


Well intuitively the function is strictly increasing initially and reaches a maximum, and clearly it will strictly decrease afterwards and go to zero pretty soon after x>k. I don't imagine it would be difficult to show with calculus, and as ceiling/floor are closest integers to the maximum one of them gives the maximum in the integers, right? And yeah I think you're right, I think a calculator will do the trick if needed :smile:
(edited 9 years ago)
Original post by jjpneed1
Well intuitively the function is strictly increasing initially and reaches a maximum, and clearly it will strictly decrease afterwards and go to zero pretty soon after x>k. I don't imagine it would be difficult to show with calculus, and as ceiling/floor are closest integers to the maximum one of them gives the maximum in the integers, right?


Yes, it is actually much easier than I thought to show that it is decreasing (the sign of the derivative is trivial to calculate). This proves in a competely rigorous way that the answer is either floor or ceiling of x/e. Still need to work out which one though.
Original post by jjpneed1
Well intuitively the function is strictly increasing initially and reaches a maximum, and clearly it will strictly decrease afterwards and go to zero pretty soon after x>k. I don't imagine it would be difficult to show with calculus, and as ceiling/floor are closest integers to the maximum one of them gives the maximum in the integers, right? And yeah I think you're right, I think a calculator will do the trick if needed :smile:


Using matlab, it is floor for x=250 and ceiling for x=1000 (as examples). There may be a pattern to which one it is but I doubt that was part of the question.
Original post by james22
Using matlab, it is floor for x=250 and ceiling for x=1000 (as examples). There may be a pattern to which one it is but I doubt that was part of the question.


I doubt it as well. The question is as follows from the '99 Trinity admissions quiz:

"I can divide the number 11 up in a number of ways, eg
11 = 9 + 2
11 = 3.1 + 4.5 + 3.2 + 0.2
but what I want is the way of splitting it up so that the product of all the numbers I split it into is as large as possible. Eg for 11 = 9 + 2 the product concerned is 9 × 2 = 18. Can you find a general way of solving this puzzle so that I could have started with numbers other than 11?"

Seeing as they'd go over the questions in the interview I suspect they'd want to see as much as we've done and perhaps they'd go over that subtlety. I'd be surprised if there was a pattern to be honest
Original post by james22
...
Original post by jjpneed1
...


Just to clarify, somebody I know IRL seized the opportunity to post the last few messages on this thread whilst my window was left open. Sorry about that!

Spoiler

(edited 9 years ago)
Original post by Khallil
Just to clarify, somebody I know IRL seized the opportunity to post the last few messages on this thread whilst my window was left open. Sorry about that!

Spoiler



Seems they were excited about it :laugh:

Posted from TSR Mobile
Original post by Khallil
Just to clarify, somebody I know IRL seized the opportunity to post the last few messages on this thread whilst my window was left open. Sorry about that!

Spoiler



Oh, I assumed that you were drunk. That makes sense though.
Problem 475*

f:  RRf:\;\mathbb{R}\to\mathbb{R}. Must there exist x,yRx,y\in\mathbb{R} s.t. f(xf(y))>x+yf(x)?f\big( x-f(y)\big)>x+yf(x)?
Original post by Lord of the Flies
...


Are you asking for us to give a counterexample for which there are no pairs x,yx,y that satisfy the inequality?

If x0x \mapsto 0, then that would imply that 0>00>0. Do you mean for the inequality to be strict?
(edited 9 years ago)
Reply 2893
34x8
Original post by Khallil
Are you asking for us to give a counterexample for which there are no pairs x,yx,y that satisfy the inequality?

If x0x \mapsto 0, then that would imply that 0>00>0. Do you mean for the inequality to be strict?


In the case f(x)=0, just consider x=-1, y=0.
Original post by Khallil
...


The question is asking whether, given any real function ff, there is a pair of real numbers that satisfies the inequality. If not we need to construct a clever counterexample, otherwise we need to prove the assertion.
Problem 476***

Prove that, for integers n2n \geq 2

e22(ne)nn!ne24(ne)n\displaystyle \dfrac{e^2}{2} \left(\dfrac{n}{e} \right)^n \leq n! \leq \dfrac{ne^2}{4}\left(\dfrac{n}{e} \right)^n
(edited 9 years ago)
Problem 477***

Evaluate the following product (note that p represents a prime number).

p(1+1p(p+1))\displaystyle \prod_p \left(1+\dfrac{1}{p(p+1)}\right)
(edited 9 years ago)
Original post by Indeterminate
Problem 477***

Evaluate the following product (note that p represents a prime number).

p(1+1p(p+1))\displaystyle \prod_p \left(1+\dfrac{1}{p(p+1)}\right)


Can you offer a hint?
Original post by jjpneed1
Can you offer a hint?


p(11p2)(1+1p(p+1))=p(11p3)\displaystyle \prod_p \left(1-\frac{1}{p^2}\right) \left(1+\frac{1}{p(p+1)}\right)=\prod_p \left(1-\frac{1}{p^3}\right)

Quick Reply

Latest