Evolvability, Complexity and Scalability of Cellular Evolutionary and Developmental Systems
Year: 2015
Type: article
Abstract: Man-made systems, such as supercomputers and software, IT-infrastructures and\nnetworks of any kind, are continuously growing in size and complexity. As\nconventional top-down engineering techniques may have reached the limit of\napplicability, biological organisms have been able to evolve increasing levels of\ncomplexity. Such inherent biological complexity may be said to be open ended or\nunbounded. This is a result of a bottom-up emergent process which produced an\nastounding diversity of living organisms with remarkable abilities, such as adaptation to\ndifferent environments or perturbations, and reproduction, being able to survive.\nEven though a lot of work has been done towards a synthesis, it is still not completely\nclear how to unleash the full potential of biological properties into artificial systems.\nThis thesis tackles the problem of better understanding the developmental process\nbetween genotype and phenotype and the evolution of complex systems made of large\nsets of elements interacting locally and giving rise to collective behaviour. In a\ntraditional Evolutionary Algorithm approach, the genotype maps to a phenotype\ndirectly, i.e. direct 1-to-1 encoding. If one wants to scale-up the phenotype complexity,\nindirect encodings, e.g. developmental or generative mappings, are a necessity. In the\nexperimental work, the chosen computational platform is Cellular Automata (CA). The\nbiological metaphor can be applied to the physical structure similarities between\nartificial cellular systems and biological multi-cellular organisms. A CA can be\nconsidered as a developing organism, where the genome specification and the gene\nregulation information control the growth and differentiation of the cells. Such a\ndynamic developmental system can show adaptation, self-modification, plasticity, and\nself-replication properties.\nIn this thesis, four challenges of designing Evolutionary and Developmental (EvoDevo)\nsystems are identified and studied further, each related to a specific research question:\nRQ1. What kind of information must be present in the genome in order to produce\ncomputation in any of the computational classes?\nRQ2. How to quantify developmental complexity, i.e. emergent phenotypic complexity?\nRQ3. Do genome parameters give any information on the evolvability of the system?\nAnd if yes, can genome information be used to guide evolutionary search in\nfavourable areas of the search space where the wanted emergent behaviour is\nmore likely to be found?\nRQ4. How can scalability of artificial EvoDevo systems be improved towards achieving\nsystems that can fully unleash their inherent complexity, potentially at the levels\nof complexity found in nature?\nThe results in this thesis show that abstract measures of phenotypic complexity may be\nsuited to characterize emergent cellular organisms. Genome information may be related\nto emergent complexity and such knowledge may be used to guide evolutionary search.\nFor scaled-up systems, it may be possible to allow indirect encodings with genome\nrepresentation growth. A framework for the evolutionary growth of genomes is\nproposed.
Author: Stefano Nichele
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2512444117