Homework
ELEG 820 - Network Optimziation
Get .m files from here
Run MakeSampleTopology
Run MakeConflictGraph
Run MakeIndeSets
With this data, find the optimal flow rates.
Note that MakeSampleTopology makes
- Destinations
– a list of destination.
- LinkPaths – a list of links crossed by the path to each
destination. LinkPaths{k} is the list of links
crossed by the path to destination(k).
Step 1. Compute the flow rate:
Given the vector Mu, compute the
data rate for each flow. There is one flow for each destination.
Step 2. Compute the link rate over
each link:
Given the flow rate to each destination, compute the data
rate over each link. We will refer to LinkRate(k) to be the required data rate over link k.
Step 3. Find the best independent
set for a give mu:
- Initialize
BestIndeBitRate = 0*Mu;
- Initialize
BestVal = -inf;
- for
each element of AllIndeSets, k, compute the sum Mu(i)*BitRate(k,i) where BitRate(k,i) is the data rate over link i
during independent set k. (Note that there is no data structure BitRate. You can make it on your own, or do this part
some other way.
Step 4. Make a new Mu:
For each element of Mu, i, set Mu(i) = Mu(i) + StepSize*(LinkRate(i) – BestIndeBitRate(i)). But if the new Mu(i)<0, then set Mu(i)=0;
Step 5. Put it all together: (See SubGradFramework.m)
- Initialize
Mu = ones(size(Links,1),1);
- Set StepSize to some small number (e.g., StepSize = 0.5).
- Initialize
AllMu=Mu; % this is
used to plot mu
- for n = 1 : Some Large Number, run steps 1-4 (in that
order).
- After
computing Mu, set AllMu
= [AllMu,Mu];
- Then
use the following code to print out AllMu
if (rem(n,1000)==0)
for i=1:length(Mu)
figure(i)
plot(AllMu(i,:))
getframe;
end
end
- After
the for n=1:some large number is finished, print All Mu
What to turn in:
- Code
- discuss
a simple example with a very small step size and explain that the mu and flow rates are correct
- Plots
for different StepSizes and different
topologies. Discuss the impact of step size.
- Extra
credit: use different bit-rates
- Also,
using class notes and the paper “Toward tractable computation of the
capacity of multihop wireless networks”
(available at http://udelmodels.eecis.udel.edu/publications1.php)
explain how the capacity optimization problems is reduced optimization of
the dual. Also, give the gradient of the dual function and related that to
the steps above.