Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Addressing high dimensional multi-objective optimization problems by coevolutionary islands with overlapping search spaces
1. Addressing high dimensional multi-
objective optimization problems by
coevolutionary islands with
overlapping search spaces
Pablo García-Sánchez, Julio Ortega,
Jesús González, Pedro A. Castillo and Juan J. Merelo
pablogarcia@ugr.es
2. Hypothesis
• Using overlapped sections of the chromosome
in a island-based co-MOEA can improve the
quality of the solutions in the same amount of
time.
3. Our approach
• Coevolutionary algorithm MOEA
• Dorronsoro et al. (2013)
• Kimovski et al. (2015): distribution and
combination
• From 8 to 128
• Overlapping sections
9. Results
• Number of island increases → quality decreases
(but better than baseline and disjoint)
• Number of island increases → overlapping
method is better → find the optimal value
• Time reduction in crossover and mutation →
higher number of generations and modifications
of the PFs
10. Conclusions and future work
• Dividing the chromosome improves all metrics
• Future: different overlapping sizes and real
problems
• Download from https://github.com/hpmoon/
hpmoon-islands