I want to optimize values of two variables in order to achieve maximum coverage in WSNs. Those two variables exist in a function that its terms changed according to acondation. What is best optimization technique that can achieve what I want ?
You can use Cross-Entropy Method. See, please, article "The Cross-Entropy Method for Continuous Multi-Extremal Optimization'
Article The Cross-Entropy Method for Continuous Multi-Extremal Optimization
Regards, Sergey.
Hi, in WSN coverage problems I suppose that your optimization problem is discrete (integer variables or mixed-integer variables) isn't it?
you may use meta-heuristic algorithms with MILP by setting your condition of variability. the program should be with multiple variables.
You could formulate the problem as a two stage optimization problem. You could try linear or non linear programming techniques. Graph theoretical approach can also be used formulating the flow in networks - A special technique known as Maximal Flow Model exists in the tool kit of operations research.
Hello Hanaa,
There is no BEST in the practical world. You are yourself using the term 'optimization'.
Therefore pl look for a method to select the most optimum tool for your problem i e to optimize values of two variables in order to achieve maximum coverage in WSNs.
KR
Virendra
Since you are considering two variables you could try classical optimization techniques.
first y you have to specify the conditions (constraints) on the objective and the objective function which expresses the coverage.
If you are able to express these in explicit form you'll easily see what type of optimization problem you have....
With only two variables you would have to resort to meta-heuristics only in the case that there is no formula expressing the coverage or if it can only be computed by a black-box algorithm - or if the properties of the functions do not allow to use general-purpose optimization methods (e.g. no convexity, but global optimum needed).
hello,
you could differentiate partially in respect to each variable and solve the equations.
Hello,
I would recommend Professor Rao on optimization techniques, it could also be downloaded via internet.
Hello dear Hanaa
If you mean that you want to solve a multi-objective problem, I recommend you to take a look at the attached paper.
I strongly recomment the paper uploaded by Amin Nobahari - in gives mathematically rigorous definitions and a rather good overview of scalarization techniques!
The overview is of course not complete (e.g. the methods by Steuer, Zionts–Wallenius, Zeleny), but all the main categories are there. And everyone could use Wikipedia to find some more....
I could not understand your problem. However, I visualize (x,y,z,p,q being terms with 0 or 1 coefficient):
{
if (a) then fa(x,y,z,p,q)
else
if(b) then fb(x,y,z,p,q)
else
if(c) then fc(x,y,z,p,q)
}
max g(fa, fb,fc) over a,b,c ? I have no knowledge of WSNs. So, it will be good that you specify the problem more mathematically.
Read this paper
https://www.researchgate.net/publication/321370891_Optimized_superpixel_and_AdaBoost_classifier_for_human_thermal_face_recognition
Read this Books
https://books.google.dz/books?id=Ez4ZAQAAIAAJ&hl=fr&source=gbs_book_other_versions
Thank you very much Saeid Parsa and Salim Aouabdi. The books you recommended are very useful.
I would use a set of evolutionary optimization algorithms and compare their performance.
- Badges
- Science topic