# Proof for Continuous function

• Mar 30th 2010, 01:37 AM
clover616
Proof for Continuous function
first, we have a function: $\displaystyle y_k$ with variables of $\displaystyle x_m$, (m=1,2,...,M)
$\displaystyle y_k = \frac{\sum_{m=1}^M x_m \alpha_{m,k}}{\sum_{m=1}^M x_m \beta_{m,k} +1}$
we know that the $\displaystyle x_m \geq 0$, $\displaystyle \alpha_{m,k}, \beta_{m,k}$ can be considered as constants.
Now, based on $\displaystyle y_k$ we have a function of $\displaystyle z$ that to the maximize the minimum of $\displaystyle y_k$,$\displaystyle (k=1,2,...,K)$ by optimally allocating the$\displaystyle x_m$ as
$\displaystyle z = \max_{x_m} \min_{k=1.\cdot,K} (y_k).$,
where $\displaystyle \sum_{m=1}^M x_m = x$,
Actually, we can alternatively express the above function of z as
$\displaystyle z=f(x),$
where the $\displaystyle x$ now becomes a variable which is the sum of the $\displaystyle x_m$ (m=1,2,...,M).
then, we want to prove $\displaystyle z=f(x),$ is continuous to x and the function f(') returns the maiximized minimum value of $\displaystyle y_k$ with an input variable of $\displaystyle x$ .

for the z=f(x), I have prove it is a strictly monotonically increasing function to x, but I failed to prove the continuity of it.

Thanks
• Mar 30th 2010, 07:48 AM
Drexel28
Quote:

Originally Posted by clover616
$\displaystyle y_k = \frac{\sum_{m=1}^M x_m \alpha_{m,k}}{\sum_{m=1}^M x_m \beta_{m,k} +1}, \sum_{m=1}^M x_m = x, x_m \geq 0, z = \max_{x_m} \min_{k=1.\cdot,K} (y_k).$
\alpha, \beta are constants

if we define a function of z=f(x), prove f(.) is continuous to x.

Thanks

I barely understand what the functions are. Could you maybe give a little background?
• Mar 30th 2010, 07:29 PM
clover616
I have modified the contents. thank!
• Mar 30th 2010, 08:04 PM
Drexel28
Quote:

Originally Posted by clover616
I have modified the contents. thank!

Please forgive my ignorance but I am still having trouble understanding. Here is the gist of what I've gotten as of now:

We have a fixed set of constants (the $\displaystyle \alpha,\beta$s) and a function $\displaystyle y:\mathbb{R}^M\to\mathbb{R}$ by a function (I understand the function itself). Clearly the order of the tuple is very relevant since for example $\displaystyle \frac{1}{100}\cdot 10+\frac{1}{10}\cdot 1\leqslant \frac{1}{10}\cdot 1+\frac{1}{100}\cdot 10$.

Now here is where I get lost. So you would agree that the above implies that we could maximize our function based on the order of our xs. But, this maximization depends on the fixed $\displaystyle \alpha,\beta$s and so really it seems that $\displaystyle z:\mathbb{R}^M\to\mathbb{R}^M$ where $\displaystyle (\alpha_1,\beta_1,\cdots)\mapsto (x_1,\cdots,x_M)$ where $\displaystyle x_1,\cdots,x_M$ are the values which maximize the function for the values of alpha and beta specified.

So, how can we fix the alpha and betas? Is this function even well defined?
• Mar 30th 2010, 08:44 PM
clover616
Quote:

Originally Posted by Drexel28
Please forgive my ignorance but I am still having trouble understanding. Here is the gist of what I've gotten as of now:

We have a fixed set of constants (the $\displaystyle \alpha,\beta$s) and a function $\displaystyle y:\mathbb{R}^M\to\mathbb{R}$ by a function (I understand the function itself). Clearly the order of the tuple is very relevant since for example $\displaystyle \frac{1}{100}\cdot 10+\frac{1}{10}\cdot 1\leqslant \frac{1}{10}\cdot 1+\frac{1}{100}\cdot 10$.

Now here is where I get lost. So you would agree that the above implies that we could maximize our function based on the order of our xs. But, this maximization depends on the fixed $\displaystyle \alpha,\beta$s and so really it seems that $\displaystyle z:\mathbb{R}^M\to\mathbb{R}^M$ where $\displaystyle (\alpha_1,\beta_1,\cdots)\mapsto (x_1,\cdots,x_M)$ where $\displaystyle x_1,\cdots,x_M$ are the values which maximize the function for the values of alpha and beta specified.

So, how can we fix the alpha and betas? Is this function even well defined?

sorry about my ambiguous definition of $\displaystyle \alpha_{m,k},\beta_{m,k}$.
actually, in the function of $\displaystyle y_k$, the $\displaystyle \alpha_{m,k},\beta_{m,k}$ are predetermined in my model regardness of the $\displaystyle x_m$ or we can say $\displaystyle \alpha_{m,k},\beta_{m,k}$ are independent to the $\displaystyle x_m$.

please have a look of my proof for the monotonicity of the funtion, then may be you can know about the problem.

In my problem , I need to focus on the continuity between $\displaystyle z$ and $\displaystyle x$, instead of that between $\displaystyle z$ and $\displaystyle x_m$ and there is no closed form solution for $\displaystyle x_m$, so this make the problem difficult.
• Mar 30th 2010, 10:35 PM
Drexel28
Quote:

Originally Posted by clover616
sorry about my ambiguous definition of $\displaystyle \alpha_{m,k},\beta_{m,k}$.
actually, in the function of $\displaystyle y_k$, the $\displaystyle \alpha_{m,k},\beta_{m,k}$ are predetermined in my model regardness of the $\displaystyle x_m$ or we can say $\displaystyle \alpha_{m,k},\beta_{m,k}$ are independent to the $\displaystyle x_m$.

please have a look of my proof for the monotonicity of the funtion, then may be you can know about the problem.

In my problem , I need to focus on the continuity between $\displaystyle z$ and $\displaystyle x$, instead of that between $\displaystyle z$ and $\displaystyle x_m$ and there is no closed form solution for $\displaystyle x_m$, so this make the problem difficult.

I really don't want to leave you hanging so I'll be straight with you. This problem seems pretty involved and I personally don't have the time to devote it the time it deserves.. I'll give it one more go a little later but don't hold your breath.
• Mar 30th 2010, 10:40 PM
clover616
Quote:

Originally Posted by Drexel28
I really don't want to leave you hanging so I'll be straight with you. This problem seems pretty involved and I personally don't have the time to devote it the time it deserves.. I'll give it one more go a little later but don't hold your breath.

Thank you very much for your kind assistance and help!(Nod)
• Mar 30th 2010, 10:42 PM
Drexel28
Quote:

Originally Posted by clover616
Thank you very much for your kind assistance and help!(Nod)

I'm sorry to say problems like this (those which take time to understand...especially because there HAS to be some kind of background to this problem) usually go unanswered on forums. People just don't have the time and aren't willing to put the effort into them. I wish you the best of luck with this problem and if you have any smaller bits of the problem that you would like help on (or another problem for that matter) please don't hesitate to re-post. Most of your question will be answered.