ABSTRACT
Differential Evolution (DE) is a powerful and versatile algorithm
for numerical optimization, but one of its downsides is its number
of parameters that need to be tuned. Multiple techniques have been
proposed to self-adapt DE’s parameters, with L-SHADE being one
of the most well established in the literature. This work presents
the A-SHADE algorithm, which modifies the population size reduction
schema of L-SHADE, and also EB-A-SHADE, which applies a
mutation strategy hybridization framework to A-SHADE. These
algorithms are applied to the CEC2013 benchmark set with 100
dimensions, and it’s shown that A-SHADE and EB-A-SHADE can
achieve competitive results.