Next: The problem of
Up: From Genetic Algorithms To
Previous: Introduction
This chapter will explain the application of the first key idea as a solution
to the local maxima problem:
- Remember the locations of local maxima and restart the optimization
program at a place distant from previously located local maxima.
I will argue that the locations of the local maxima discovered so far can be
used to guide further search in the space. The first section will introduce
the problem of local maxima, and describe the initial attempts to solve it.
The second section is a discussion on the optimum population size and the
possible implementation on parallel computers. The third section will explain
how to use the local maxima, and introduce the top level procedure for our
method. The fourth section presents several solutions to the problem of
finding points with high diversity. Finally the fifth section will conclude
the chapter by a discussion of two other approaches inspired by natural
phenomena, Hillis's idea of ``co-evolving parasites'', and the simulated
annealing algorithm.
Deniz Yuret
Tue Apr 1 21:38:29 EST 1997