DCMA, yet another derandomization in covariance matrix adaptation
Olivier Teytaud and Justin Bedo
In: Gecco 2007, 2007, London.
In a preliminary part of this paper, we analyze the neces-
sity of randomness in evolution strategies. We conclude to
the necessity of ”continuous”-randomness, but with a much
more limited use of randomness than what is commonly used
in evolution strategies. We then apply these results to CMA-
ES, a famous evolution strategy already based on the idea
of derandomization, which uses random independent Gaus-
sian mutations. We here replace these random independent
Gaussian mutations by a quasi-random sample. The mod-
iﬁcation is very easy to do, the modiﬁed algorithm is com-
putationally more eﬃcient and its convergence is faster in
terms of the number of iterates for a given precision.