Acquiring evolvability through adaptive representations

Abstract

Adaptive representations allow evolution to explore the space of phenotypes by choosing the most suitable set of genotypic parameters. Although such an approach is believed to be efficient on complex problems, few empirical studies have been conducted in such domains. In this paper, three neural network representations, a direct encoding, a complexifying encoding, and an implicit encoding capable of adapting the genotype-phenotype mapping are compared on Nothello, a complex game playing domain from the AAAI General Game Playing Competition. Implicit encoding makes the search more efficient and uses several times fewer parameters. Random mutation leads to highly structured phenotypic variation that is acquired during the course of evolution rather than built into the representation itself. Thus, adaptive representations learn to become evolvable, and furthermore do so in a way that makes search efficient on difficult coevolutionary problems.

Evolving neural fields for problems with large input and output spaces

“…New connections are established either within a given module or between a given module and m 0 . In yet another approach, sets of rules are evolved with NEAT-like speciation that implicitly define a neural network (Reisinger and Miikkulainen, 2007).…”

Brief Review of Techniques Used to Develop Adaptive Evolutionary Algorithms

“…Other work is Reisinger and Miikkulainen [40] where they used a new class of representations for real valued parameters called Center of Mass Encoding (CoME). CoME is based on variable length strings and it allows the choice of redundancy degree of the genotype-phenotype map and the choice of redundancy distribution for the space of phenotypes.…”

Comparing the Evolvability of Generative Encoding Schemes

“…Inspired by the evolvability of biological systems, researchers in EC have abstracted the underlying developmental processes, to formulate generative genotype-phenotype maps for artificial systems (e.g., Stanley (2007)). The resultant generative encodings frequently outperform traditional direct encodings for various application problems such as, designing 3D objects (e.g., Hornby (2005)), game playing (e.g., Reisinger and Miikkulainen (2007); Gauci and Stanley (2010)), pattern matching (e.g., Clune et al (2011)), and robot locomotion (e.g., Hornby and Pollack (2002); Seys and Beer (2007)). Furthermore, the higher evolvability provided by generative encodings is often considered as the reason for the observed differences in performance, consequent to their capability to reuse parts of the genotype to affect different phenotypes, scale well to large phenotypic spaces, and generate modular architectures (Stanley and Miikkulainen, 2003).…”

“…Most studies estimate evolvability either as, (i) the proportion of genetic mutations that are beneficial to an individual, irrespective of the phenotypic novelty of the resultant offspring (e.g., Hornby et al (2003); Reisinger and Miikkulainen (2007)), or as (ii) the range and diversity of the phenotypic variants resulting from genetic change (Lehman and Stanley, 2011;Reisinger et al, 2005;Lehman and Stanley, 2013), usually without considering the deleteriousness of the change. Importantly, both these estimates when considered alone do not discount for mutations that, (i) generate very diverse phenotypes but prove lethal to an organism, and (ii) result in minor improvements to a phenotype, but are unable to generate novelty.…”