Incorporating a priori knowledge in probabilistic-model based optimization

Abstract

Recent studies have examined the effectiveness of using probabilistic models to guide the sample generation process for searching high dimensional spaces. Building complex dependency networks that can account for the interactions between parameters are often used; however, they may necessitate enormous amounts of sampling. In this chapter, we demonstrate how a priori knowledge of parameter dependencies, even incomplete knowledge, can be incorporated to efficiently obtain accurate models that account for parameter interdependencies. This is achieved by effectively putting priors on the network structures that are created. These more accurate models yield improved results when used to guide the sample generation process for search. We demonstrate the results on a variety of graph coloring problems, and examine the benefits of a priori knowledge as problem difficulty increases. Recent studies have examined the effectiveness of using probabilistic models to guide the sample generation process for searching high dimensional spaces. Building complex dependency networks that can account for the interactions between parameters are often used; however, they may necessitate enormous amounts of sampling. In this chapter, we demonstrate how a priori knowledge of parameter dependencies, even incomplete knowledge, can be incorporated to efficiently obtain accurate models that account for parameter interdependencies. This is achieved by effectively putting priors on the network structures that are created. These more accurate models yield improved results when used to guide the sample generation process for search. We demonstrate the results on a variety of graph coloring problems, and examine the benefits of a priori knowledge as problem difficulty increases.Recent studies have examined the effectiveness of using probabilistic models to guide the sample generation process for searching high dimensional spaces. Building complex dependency networks that can account for the interactions between parameters are often used; however, they may necessitate enormous amounts of sampling. In this chapter, we demonstrate how a priori knowledge of parameter dependencies, even incomplete knowledge, can be incorporated to efficiently obtain accurate models that account for parameter interdependencies. This is achieved by effectively putting priors on the network structures that are created. These more accurate models yield improved results when used to guide the sample generation process for search. We demonstrate the results on a variety of graph coloring problems, and examine the benefits of a priori knowledge as problem difficulty increases. © Springer-Verlag Berlin Heidelberg 2006.