# 3.5 Constrained optimization with inequality constraints  (Page 2/2)

With some additional assumptions, it can be shown that the KKT conditions can find a global minimizer.

Definition 3 A function $f$ is said to be affine over $\Omega$ if $f\left({\sum }_{i}^{n}{a}_{i}{x}_{i}\right)={\sum }_{1}^{n}{a}_{i}f\left({x}_{i}\right)$ for all ${x}_{1},...,{x}_{n}\in \Omega$ and all weights $\left\{{a}_{i}\right\}$ obeying ${\sum }_{i}^{n}{a}_{i}=1$ .

Theorem 2 (Karush-Kuhn-Tucker Sufficient Conditions) If $f$ and ${h}_{j}$ , $j=1,...,m$ are convex functions and ${g}_{i}$ , $i=1,...,n$ are affine functions, and if the KKT condition are satisfied at a feasible point ${x}_{0}\in \Omega$ then ${x}_{0}$ is a global minimizer of $f$ over $\Omega$ .

Fix ${x}_{1}\in \Omega$ let $d={x}_{1}-{x}_{0}$ . Define a functional $x\left(t\right)=t{x}_{1}+\left(1-t\right){x}_{0}={x}_{0}+td$ over $t\in \left[0,1\right]$ . Then, define the constraints limited over the set of points $x\left(t\right)$ :

$\begin{array}{cc}\hfill {G}_{i}\left(t\right)& ={g}_{i}\left(x\left(t\right)\right)={g}_{i}\left(t{x}_{1}+\left(1-t\right){x}_{0}\right)=t\phantom{\rule{0.277778em}{0ex}}g\left({x}_{i}\right)+\left(1-t\right)g\left({x}_{0}\right)=0,\hfill \\ \hfill {H}_{j}\left(t\right)& ={h}_{j}\left(x\left(t\right)\right)={h}_{j}\left(t{x}_{1}+\left(1-t\right){x}_{0}\right)\le t{h}_{j}\left({x}_{1}\right)+\left(1-t\right){h}_{j}\left({x}_{0}\right)\le 0;\hfill \end{array}$

Therefore, all points $x\left(t\right)\in \Omega$ are feasible. Furthermore, note that ${H}_{j}\left(0\right)={h}_{j}\left({x}_{0}\right)=0\ge {H}_{j}\left(t\right)={h}_{j}\left({x}_{t}\right)$ if $j\in J\left({x}_{0}\right)$ . Now, we compute the derivatives of these two functions with respect to $t$ :

$\frac{\partial G}{\partial t}=0=\partial {g}_{i}\left({x}_{0},d\right)=⟨\nabla {g}_{i}\left({x}_{0}\right),d⟩,$

and for $j\in J\left({x}_{0}\right)$ ,

$0\phantom{\rule{0.277778em}{0ex}}\ge \phantom{\rule{0.277778em}{0ex}}\frac{\partial {H}_{j}}{\partial t}=\partial {h}_{j}\left({x}_{0},d\right)=⟨\nabla {h}_{j}\left({x}_{0}\right),d⟩.$

Now consider the function $F\left(t\right)=f\left(x\left(t\right)\right)$ : its derivative is given by

$\frac{\partial F}{\partial t}=\partial f\left({x}_{0},d\right)=⟨\nabla f\left({x}_{0}\right),d⟩=-\sum _{i=1}^{n}⟨\nabla {g}_{i},d⟩{x}_{i}-\sum _{i=1}^{n}{\mu }_{j}⟨\nabla {h}_{j},d⟩\ge 0,$

where we use the third KKT condition. Since $f\left(x\right)$ is convex and $x\left(t\right)$ is affine, then $F\left(t\right)=f\left(x\left(t\right)\right)$ is convex in $t\in \left[0,1\right]$ . Thus $\frac{\partial F}{\partial t}$ is nondecreasing and $\frac{\partial F\left(t\right)}{\partial t}\ge \frac{\partial F\left(0\right)}{\partial t}\ge 0$ for $t\in \left[0,1\right]$ . Thus, $F\left(1\right)\ge F\left(0\right)$ or $f\left({x}_{1}\right)\ge f\left({x}_{0}\right)$ . Since ${x}_{1}$ was arbitrary, ${x}_{0}$ is a global minimum of $f$ on $\Omega$ .

Example 2 (Channel Capacity) The Shannon capacity of an additive white Gaussian noise channel is given by $C=\frac{1}{2}{log}_{2}\left(1+\frac{P}{N}\right)$ , where $P$ is the transmitted signal power and $N$ is the noise variance. Assume that $n$ channels are available with a total transmission power ${P}_{T}={\sum }_{i=1}^{n}{P}_{i}$ available among the channels, where ${P}_{i}$ denotes the power in the ${i}^{th}$ channel. We wish to assign a power profile $P={\left[{P}_{1},...,{P}_{n}\right]}^{T}$ that maximizes the total capacity for the set of channels

$C\left(P\right)=\sum _{i=1}^{n}C\left({P}_{i}\right)=\sum _{i=1}^{n}\frac{1}{2}{log}_{2}\left(1,+,\frac{{P}_{i}}{{N}_{i}}\right),$

where ${N}_{i}$ represents the variance of the noise in the ${i}^{th}$ channel.

To solve the problem, we set up an objective function to be minimized

$f\left(P\right)=-C\left(P\right)=-\sum _{i=1}^{n}C\left({P}_{i}\right)=-\sum _{i=1}^{n}\frac{1}{2}{log}_{2}\left(1+\frac{{P}_{i}}{{N}_{i}}\right)$

and also set up the constraints

$\begin{array}{cc}\hfill g\left(P\right)& =\sum _{i=1}^{n}{P}_{i}-{P}_{T}={\mathbf{1}}^{T}P-{P}_{T},\hfill \\ \hfill {h}_{i}\left(P\right)& =-{P}_{i}=-{e}_{i}^{T}P,\phantom{\rule{3.33333pt}{0ex}}i=1,...,n,\hfill \end{array}$

as the values of the powers must be nonnegative. We start by computing the gradients of these functions: for $f$ , we must compute the directional derivative

$\begin{array}{cc}\hfill \delta f\left(p;h\right)& =\frac{\partial }{\partial \alpha }{\left(f\left(p+\alpha h\right)\right)|}_{\alpha =0}={\left(\left(-,\sum _{i=1}^{n},\frac{{h}_{i}/{N}_{i}}{2\left(ln2\right)\left(1,+,\frac{{P}_{i}}{{N}_{i}},+,\alpha ,\frac{{h}_{i}}{{N}_{i}}\right)}\right)|}_{\alpha =0},\hfill \\ & =-\sum _{i=1}^{n}\frac{{h}_{i}}{2{N}_{i}\left(ln2\right)\left(1,+,\frac{{P}_{i}}{{N}_{i}}\right)}=-\sum _{i=1}^{n}\frac{{h}_{i}}{2\left(ln2\right)\left({N}_{i}+{P}_{i}\right)}=⟨\nabla f\left(p\right),h⟩,\hfill \end{array}$

where the gradient has entries ${\left(\nabla f\left(p\right)\right)}_{i}=-{\left(2\left(ln2\right)\left({N}_{i}+{P}_{i}\right)\right)}^{-1}$ .

For the constraints, it is straightforward to see that $\nabla g\left(P\right)=\mathbf{1}$ and $\nabla {h}_{i}\left(P\right)=-{e}_{i}$ , $i=1,...,n$ .

We begin by assuming that the solution ${P}^{*}$ is a regular point. Then the KKT conditions give that for some $\lambda$ and nonnegative ${\mu }_{1},...,{\mu }_{m}$ we must have

$\begin{array}{cc}\hfill \sum _{i=1}^{n}{\mu }_{i}{P}_{i}^{*}& =0,\hfill \\ \hfill -\frac{1}{2\left(ln2\right)\left({N}_{i}+{P}_{i}^{*}\right)}+\lambda -{\mu }_{i}& =0,\phantom{\rule{3.33333pt}{0ex}}i=1,...,n.\hfill \end{array}$

The second set of constraints can be written as

$\begin{array}{cc}\hfill \frac{1}{2\left(ln2\right)\left({N}_{i}+{P}_{i}^{*}\right)}& =\lambda -{\mu }_{i},\hfill \\ \hfill {N}_{i}+{P}_{i}^{*}& =\frac{1}{2\left(ln2\right)\left(\lambda -{\mu }_{i}\right)}.\hfill \end{array}$

Consider each inequality constraint ${h}_{i}$ .

• If ${h}_{i}$ is inactive, then ${P}_{i}^{*}>0$ and ${\mu }_{i}=0$ . Then,
$\begin{array}{cc}\hfill {N}_{i}+{P}_{i}^{*}& =\frac{1}{2\lambda \left(ln2\right)},\hfill \\ \hfill {P}_{i}^{*}& =\frac{1}{2\lambda \left(ln2\right)}-{N}_{i}>0.\hfill \end{array}$
• If ${h}_{i}$ is active, then ${P}_{i}^{*}=0$ and so
$\begin{array}{cc}\hfill {N}_{i}& =\frac{1}{2\left(ln2\right)\left(\lambda -{\mu }_{i}\right)},\hfill \\ \hfill {\mu }_{i}& =\lambda -\frac{1}{2{N}_{i}\left(ln2\right)}\ge 0,\hfill \\ \hfill \frac{1}{2{N}_{i}\left(ln2\right)}& \le \lambda ,\hfill \\ \hfill \frac{1}{2\lambda \left(ln2\right)}& \le {N}_{i}.\hfill \end{array}$

To simplify, write $r=\frac{1}{2\lambda \left(ln2\right)}$ ; then, we have two possibilities for each channel $i$ from above:

• If $r-{N}_{i}>0$ (i.e., if ${N}_{i} ), then ${P}_{i}^{*}=r-{N}_{i}$ .
• If $r-{N}_{i}\le 0$ (i.e., if $r\le {N}_{i}$ ) then ${P}_{i}^{*}=0$ .

Thus the power is allocated among the channels using the formula ${P}_{i}^{*}=max\left(0,r-{N}_{i}\right)$ , and the value of $r$ is chosen so that the total power constraints is met:

$\sum _{i=1}^{n}max\left(0,r-{N}_{i}\right)={P}_{T}.$

This is the famous water-filling solution to the multiple channel capacity problem, illustrated in [link] .

Application of nanotechnology in medicine
what is variations in raman spectra for nanomaterials
I only see partial conversation and what's the question here!
what about nanotechnology for water purification
please someone correct me if I'm wrong but I think one can use nanoparticles, specially silver nanoparticles for water treatment.
Damian
yes that's correct
Professor
I think
Professor
what is the stm
is there industrial application of fullrenes. What is the method to prepare fullrene on large scale.?
Rafiq
industrial application...? mmm I think on the medical side as drug carrier, but you should go deeper on your research, I may be wrong
Damian
How we are making nano material?
what is a peer
What is meant by 'nano scale'?
What is STMs full form?
LITNING
scanning tunneling microscope
Sahil
how nano science is used for hydrophobicity
Santosh
Do u think that Graphene and Fullrene fiber can be used to make Air Plane body structure the lightest and strongest. Rafiq
Rafiq
what is differents between GO and RGO?
Mahi
what is simplest way to understand the applications of nano robots used to detect the cancer affected cell of human body.? How this robot is carried to required site of body cell.? what will be the carrier material and how can be detected that correct delivery of drug is done Rafiq
Rafiq
if virus is killing to make ARTIFICIAL DNA OF GRAPHENE FOR KILLED THE VIRUS .THIS IS OUR ASSUMPTION
Anam
analytical skills graphene is prepared to kill any type viruses .
Anam
what is Nano technology ?
write examples of Nano molecule?
Bob
The nanotechnology is as new science, to scale nanometric
brayan
nanotechnology is the study, desing, synthesis, manipulation and application of materials and functional systems through control of matter at nanoscale
Damian
Is there any normative that regulates the use of silver nanoparticles?
what king of growth are you checking .?
Renato
What fields keep nano created devices from performing or assimulating ? Magnetic fields ? Are do they assimilate ?
why we need to study biomolecules, molecular biology in nanotechnology?
?
Kyle
yes I'm doing my masters in nanotechnology, we are being studying all these domains as well..
why?
what school?
Kyle
biomolecules are e building blocks of every organics and inorganic materials.
Joe
anyone know any internet site where one can find nanotechnology papers?
research.net
kanaga
sciencedirect big data base
Ernesto
Introduction about quantum dots in nanotechnology
hi
Loga
what does nano mean?
nano basically means 10^(-9). nanometer is a unit to measure length.
Bharti
Got questions? Join the online conversation and get instant answers!