hi dear researchers.
consider two design.
1)optimal design without considering robustness(optimization design).
2)optimal design with considering robustness(robust optimization design ).
i read in a text that robustness and performance are in conflict.
now it is my question.
is it possible that robust design has better performance( as well as better robustness) than optimization design or not?
in other hand,
is it possible robust design has both better performance and robustness?(robust design has lower objective function than optimal design in minimization problem?)
Dear Mr. Setayandeh,
Sorry, you should give more information. Many people say that everything is possible. Others say that everything in this life is a trade-off among different objectives.
My advice. You should think carefully what is the most important objective of your quest (optimality or robustness). Actually, this should be the first step in any investigation, defining its objective. This will help you to choose one option or another. As users, statistics is a tool like a knife, a screwdriver, a vehicle or even a pair of shoes. You would choose a SUV, a Ferrari or a truck depending on your objective.
Following the comparison, humans typically use tools in series. The same can be done with statistics, first use one design for the most important objective. Then choose another one for the second. It could be easier to choose the next step with the information gained after the first stage. After all, statistics is a tool to gain information about systems.
Hope it helps,
robert
Hi Setayandeh,
All optimal solutions need not be robust ,which means a slight perturbation in any input parameter of a less robust optimal solution will lead to huge performance loss.This happens sometimes with the optimally arrived configurations using stochastic methods.While you arrive at any configuration or design,you must consider the robustness for sure because nothing can be perfectly fabricated or no system remains intact through out its life.So a stable/robust design is more acceptable from an engineering point of view.
To check the robustness of the function you must analyse the performance of the function in the vicinity of its optimum.
regards
Rohit
Hi, It is possible, see the below references
[] Robust Worst-Case Interference Control in Underlay Cognitive Radio Networks
Parsaeefard, S.; Sharafat, A.R.
Vehicular Technology, IEEE Transactions on
[] Robust Power Control for Heterogeneous Users in Shared Unlicensed Bands
Parsaeefard, S.; van der Schaar, M.; Sharafat, A.R.
Wireless Communications, IEEE Transactions on
[]
Robust Additively Coupled Games in the Presence of Bounded Uncertainty in Communication Networks
Parsaeefard, S.; Sharafat, A.R.; van der Schaar, M.
In the usual case, that is, when you get the optimal solution of the problem in two deterministic and robust optimization cases, the solution of the deterministic model will definitely have a lower value. But suppose that you have designed a supply chain network using deterministic data. Now, if uncertainties occur, the network that you have designed using certain data will definitely increase the costs. I suggest you read the numerical results part of my paper, which has been properly analyzed in this field:
https://www.tandfonline.com/doi/full/10.1080/00207543.2021.1953179
- Badges
- Science topic