Scatter Search

Scatter search This chapter discusses the principles and foundations behind scatter search and its application to the problem of training neural networks. Scatter search is an evolutionary method that has been successfully applied to a wide array of hard optimization problems. Scatter search constructs new trial solutions by combining so-called reference solutions and employing strategic designs that exploit context knowledge. In contrast to other evolutionary methods like genetic algorithms, scatter search is founded on the premise that systematic designs and methods for creating new solutions afford significant benefits beyond those derived from recourse to randomization. Our implementation goal is to create a combination of the five elements in the scatter search methodology that proves effective when searching for optimal weight values in a multilayer neural network. Through experimentation, we show that our instantiation of scatter search can compete with the best-known training algorithms in terms of training quality while keeping the computational effort at a reasonable level. (Source: http://plato.asu.edu)