Dear researchers,
First of all this is hard question to put forward for me, but let me try.
In our work, GA is used to optimize fuzzy membership functions for sink mobility. It is to determine after how many events subsequent sink re-location is considered. This method along with re-clustering (after how many nodes depletion next phase of re-clustering takes place) help achieve good extension in network lifetime.
However, point is there can be pros and cons for using same simulator for "GA optimization" and use it later for "experiments".
-One aspect is that we are just "copying" the performance obtained during optimization of fuzzy membership functions to actual experiments which use it. Therefore, for two things, different simulators should be used.
(In this regard, question is in how two simulators should be different?)
-Second aspect is that it is ideal to determine optimized something and use it to get best performance for experiments, however it has problem as described in point one "of copying performance".
After some reading and thought, since, we know fitness function is component in GA which interact with simulation, there should be some mechanism to measure performance within fitness function. So, what aspect of learning in GA fuzzy optimization, we can use later in experiments that we are not just copying performance.
It would be great to hear some feedback, in order to determine how to respond it.
Waiting for your good response and any question related to response.
Thanks, Akash~
It seems a good combination for optimization and experimental evaluation.
let me bring this to you from another perspective:
Using same simulators for optimization and experimental evaluations have merits. However, for the purpose of robustness, diversity, and removing bias characteristics experimental evaluation could be performed in the different simulator. In the same simulator, since the behavior of events for the time of optimization and later experimental evaluation is different performance is not duplicated. Some may argue layout should be same for fairness and some argue should be changed for to avoid bias and test robustness.
What I need to know is pros and cons of using same or different simulators for optimization and experimental evaluation.
Using several simulators is always a good idea because they all include different simulations variables and metrics. Nevertheless, some softwares such as MATLAB offer many tools and sometimes you just get all you want in a single package. Besides, the gain of using alternative software is often reduced when compared to the time spent on research and learning.
Thanks Jose,
Is there any advantages in using identical simulators?
regards
- Badges
- Science topic