A brief survey of Genetic Algorithms as applied to non-linear optimisation ∗

Some non-linear optimisation problems are difficult to solve by con-ventional hill-climbing methods, as they get “stuck ” in local minima. In an attempt to avoid this problem, a number of global maximi-sation/minimisation strategies have been developed using stochastic methods. One such method which has gained some publicity of late is Genetic Algorithms. This paper gives an overview of how genetic algorithms work. It includes a sample program for maximising any n-dimensional function along with detailed commentary on how it works and its performance on a few sample problems. Modified forms of the standard genetic algorithm, known as Niche methods are used to locate multiple optima. A brief discussion of this extension is made. 1 The need for global maximisation A common mathematical problem to be solved in many disciplines is that of minimising or maximising a given function. In fact the two problems ar