On identifying global optima in cooperative coevolution

Abstract

When applied to optimization problems, Cooperative Coevolutionary Algorithms (CCEA) have been observed to exhibit a behavior called relative overgeneralization. Roughly, they tend to identify local optima with large basins of attraction which may or may not correspond to global optima. A question which arises is whether one can modify the algorithm to promote the discovery of global optima. We argue that a mechanism from Pareto coevolution can achieve this end. We observe that in CCEAs candidate individuals from one population are used as tests or measurements of individuals in other populations; by treating individuals as tests in this way, a finer-grained comparison can be made among candidate individuals. This finer-grained view permits an algorithm to see when two candidates are differently capable, even when one's evident value is higher than the other's. By modifying an existing CCEA to compare individuals using Pareto dominance we have produced an algorithm which reliably finds global optima. We demonstrate the algorithm on two Maximum of Two Quadratics problems and discuss why it works.

Cooperative Coevolution for Agrifood Process Modeling

“…For example, a naive way of storing the conditional probabilities of 10 binary variables as a table requires storage space for 2 10 = 1024 values. If the local distributions of no variable depends on more than 3 parent variables, the Bayesian network representation only needs to store at most 10 * 2 3 = 80 values.…”

“…Individuals of a subpopulation are evaluated by aggregation with individuals of other subpopulations. Multi-species cooperative co-evolution has been applied to various problems [43,55,54,22,36,66], including learning problems [8], and some theoretical analyses have been recently proposed, see [48,10,52], or [65] for an analysis considering a relationship between cooperative co-evolution and evolutionary game theory.…”

Initial Results from Co-operative Co-evolution for Automated Platformer Design

“…CCE has unique problems associated with it that affects its ability to locate global optima, which we are yet to investigate in ACCME. Such problems are discussed in [17] and some solutions proposed in [18]. We aim to apply these ideas to ACCME in the hope that it will improve the process.…”

Reference sharing: a new collaboration model for cooperative coevolution

“…Most of the current methods implement a natural decomposition where each component represents one or multi-dimensions of the optimized structure. The one dimension could be a single variable in a function optimization (Bucci and Pollack 2005;Potter and De Jong 1994;Potter 1997), or a hidden neuron in an evolved artificial neural network (Gomez 2003;Moriarty and Miikkulainen 1997). 2.…”

“…The first case decomposes, usually manually, a problem before starting the evolutionary process, and it does not alter the decomposed components afterwards (Bucci and Pollack 2005;Panait and Luke 2005;Potter and Jong 2000). The second case predecomposes a problem at the beginning, but components are able to be self-adaptively tuned to proper interaction levels during the evolutionary process (Ray and Yao 2009;Weicker and Weicker 1999;Yang et al 2008a, b;Omidvar et al 2014).…”