a new strategy fitness have proposed by Xiufen Zou In article "A High Performance Multi-objective Evolutionary Algorithm Based on the Principles of Thermodynamics" and some paper referenced to it, actually I am working in multi objective optimization and I want to use that fitness.
The best solution has the lowest fi, and the worst one has the largest fi , the fitness is assigned as the inverse of fi:
Here, Si = − pT (i) log pT (i), where T > 0 is temperature.
p (i) = (1/Z) exp ( − ri /T) is the Gibbs distribution.
z= Σexp (-ri) /T
N1 i is known as the partition function, N is the
population size . Here, ri is the Pareto-rank of solution i and cdi
is the crowding distance of solution i computed by density estimation
technique suggested in ( i is index).
I coded them in matlab . but every time results are in probability of variable zero for solution with crowding distance infinity and other solutions are 0.02........
My results never cover the pareto front optimal, It means some solutions far away from pareto front optimal. article and results is shown as bellow.
I am working on "NSABC:Non-dominated sorting based multi-objective artificial" which have proposed by Avadh Kishor, Pramod KumarSingh and JayPrakash
http://mayankagr.in/images/matlab_tutorial.pdf
https://www.tutorialspoint.com/matlab/matlab_pdf_version.htm
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Algorithms