SWAY, the Sampling WAY, is baseline optimizer for multi-objective optimization problems, especially for the search-based software engineering problems. SWAY does not have the crossover and mutation in the standard evolutionary algorithm. Instead, it samples from a large candidate pool.
@article{chen2016sampling,
title={Is" Sampling" better than" Evolution" for Search-based Software Engineering?},
author={Chen, Jianfeng and Nair, Vivek and Krishna, Rahul and Menzies, Tim},
journal={arXiv preprint arXiv:1608.07617},
year={2016}
}
- Clone this repo
- Define your own problem!
- Create the initial candidate pool.
- Use SWAY
In the following example, we create our own problem Networking with 4-dimension objectives QoS and use SWAY to search for optimum.
SWAY accepts the problems defined in DEAP
from Algorithms import sway_continous, sway_discrete
from deap import base, creator, tools
creator.create('Qos', base.Fitness, weights=(-1.0, -1.0, -1.0, -1.0)) # note-a
creator.create('Networking', array.array, typecode='d', fitness=creator.Qos) # note-b
def modelEvaluation(ind):
ind.fitness.values = (1.1, 2.2, 3.3, 4.4) # note-c
init = ... # note-d
res = sway_continous.optimize(init, modelEvaluation)
for i in res:
print(i.fitness.values)a) what is Fitneess? weights?
b) what is typecode?
c) The objective of individual is assigned "in-place".
The decision of the ind can be accessed by ind[0], ind[1],...
d) Creating the candidates pool and put all into a list. A candidate can be created by Networking([1,2,3,4,5,6])