Here's the question you clicked on:
FoolForMath
Just another super easy problem, Let \(x, y, z \in \mathbb{R}^+\) such that \(x +y +z =\sqrt{3} \) . Prove that the maximum value of \(\large \frac x{\sqrt{x^2+1}} +\frac y{\sqrt{y^2+1}} +\frac z{\sqrt{z^2+1}}\) is \(\frac 32 \)
And yes we don't need to use Lagrange multiplier. ( http://en.wikipedia.org/wiki/Lagrange_multiplier ) otherwise it will be boring.
we could write:\[\frac{x}{\sqrt{x^2+1}}=\sqrt{\frac{x^2}{x^2+1}}=\sqrt{\frac{x^2+1-1}{x^2+1}}=\sqrt{1-\frac{1}{x^2+1}}\]so then we want to maximize:\[\sqrt{1-\frac{1}{x^2+1}}+\sqrt{1-\frac{1}{y^2+1}}+\sqrt{1-\frac{1}{z^2+1}}\]
which means we need to maximize each of the denominators
which gives me the /gut/ feeling that x=y=z is the solution since we have a constraint on the sum of x, y and z
but I cannot prove my /gut/ is correct :)
The reason why I liked this problem is the fact that it uses an inequality which is very useful in general.
is it worth pursuing: let x = a\(\sqrt{3}\) let y = b\(\sqrt{3}\) let z = c\(\sqrt{3}\) so we have: a + b + c = 1
AMGM? Cauchy-Schwarz? Chebyshev? (These are the only ones I know.)
I guess thats a no to both blockcolder and myself :(
I like the way asnaseer approaches any problem, always starting from scratch :)
I was talking about Jensen's Inequality ( http://en.wikipedia.org/wiki/Jensen's_inequality )
\[x \sqrt{y ^{2}+1}\sqrt{z ^{2}+1} = x \sqrt{(y ^{2}+1)(z ^{2}+1})\]
\[x \sqrt{y ^{2}+y ^{2}z ^{2}+z ^{2}+y ^{2}z ^{2}}\]
Your a liar, This is hard
Oh, jensen's. Therefore, asnaseer's approach is worth looking at.
x(y+z) do the same thing for all of them
If you know jensen's this is just there lines.
never had the pleasure of meeting him :)
yeah you guys can just leave me do this alone
You were here at 1906 ? :P
of course - weren't you?
19:06 == 6 minutes past 7pm
xy + zy + yx + zy + zx + zy 2zx + 2xy + 2zy 2y(x+z) + 2zx x+z = sqrt 3 - y
\[2({\sqrt{3}-y}) + 2x(\sqrt{3} - x -y)\]
\[2\sqrt{3} - 2y + 2{\sqrt{3}}x - 2x ^{2} - 2xy\]
differentiate to find max value
Jensen's goes like this, right? \[\lambda_1+\lambda_2+\cdots+\lambda_n=1\\ f''(x)>0\\ \therefore f(\lambda_1x_1+\lambda_2x_2+\cdots+\lambda_nx_n)\leq\lambda_1f(x_1)+\lambda_2f(x_2)+\cdots+\lambda_nf(x_n) \]
I /think/ you are correct blockcolder from what I've just read about this method, but I don't understand it well enough to apply it here.
This is actually concave. f''(x)<0
I /think/ the sum pf your lambda's equate equate to: a+ b + c = 1
I was successful in avoiding the Hessian by keeping things only to one variable.
could we divide everything by \(\sqrt{3}\) to get:\[x'+y'+z'=1\]where:\[x'=\frac{x}{\sqrt{3}}\]etc
so then we need to maximize:\[\frac{x'}{\sqrt{x'^2+3}}+...\]
sorry, that should be - we need to maximize:\[\frac{x'\sqrt{3}}{\sqrt{3x'^2+1}}+...\]
so now the x', y' and z' correspond to your lambda's blockcolder
I'm probably way off the mark as its very late here and I need to get some sleep. good like with this guys. I'll check-in sometime tomorrow (or should I say later today) to how how this was solved.