RASH: A Self-adaptive Random Search Method

Mauro Brunato, Roberto Battiti

Abstract:

This paper presents
an adaptive stochastic search algorithm for the optimization of
functions of continuous variables where the only hypothesis is the pointwise
computability of the function.

The main design criterion of the proposed scheme consists
of the adaptation of a search region by an affine transformation
which takes into account the local knowledge derived from trial
points generated with uniform probability.
Heuristic adaptation of the step size and direction allows
the largest possible movement per function evaluation.
Our results show that the proposed technique is, in spite of its simplicity, a
promising building block to consider for the development of more
complex optimization algorithms, particularly in those cases where the objective function evaluation
is expensive.