Optimization

139

5

Guided Random Search Methods

5.1 Introduction

The solution techniques for unconstrained optimization problems that have

been described in earlier chapters invariably use the gradient information

to locate the optimum. Such methods, as we have seen, require the objec-

tive function to be continuous and differentiable, and the optimal solution

depends on the chosen initial conditions. These methods are not efcient in

handling discrete variables and are more likely to stay at a local optimum for

a multimodal objective function. Gradient-based methods often have to be

restarted to ensure that the local optimum reached is indeed the global one. ...