This paper introduces a continuous model for Multi-cellular Developmental Design. The cells are fixed on a 2D grid and exchange ”chemicals ” with their neighbors during the growth process. The quantity of chemicals that a cell produces, as well as the differentiation value of the cell in the phenotype, are controlled by a Neural Network (the genotype) that takes as inputs the chemicals produced by the neighboring cells at the previous time step. In the proposed model, the number of iterations of the growth process is not pre-determined, but emerges during evolution: only organisms for which the growth process stabilizes give a phenotype (the stable state), others are declared nonviable. The optimization of the controller is done using the NEAT algorithm, that optimizes both the topology and the weights of the Neural Networks. Though each cell only receives local information from its neighbors, the experimental results of the proposed approach on the ’flags ’ problems (the phenotype must match a given 2D pattern) are almost as good as those of a direct regression approach using the same model with global information. Moreover, the resulting multi-cellular organisms exhibit almost perfect self-healing characteristics.

Abstract. Echo State Networks (ESN) have demonstrated their efficiency in supervised learning of time series: a ”reservoir ” of neurons provide a set of dynamical systems that can be linearly combined to match the target dynamics, using a simple quadratic optimisation algorithm to tune the few free parameters. In an unsupervised learning context, however, another optimiser is needed. In this paper, an adaptive (1+1)-Evolution Strategy is used to optimise an ESN to tackle the ”flag ” problem, a classical benchmark from multi-cellular artificial embryogeny: the genotype is the cell controller of a Continuous Cellular Automata, and the phenotype, the image that corresponds to the fixedpoint of the resulting dynamical system, must match a given 2D pattern. This approach is able to provide excellent results with few evaluations, and favourably compares to that using the NEAT algorithm (a stateof-the-art neuro-evolution method) to evolve the cell controllers. Some characteristics of the fitness landscape of the ESN-based method are also investigated. 1

V´ypočetní development pˇredstavuje obsáhlou podoblast evolučního návrhu. Obecně je development chápán jako pˇrídavn´y mechanismus evolučního algoritmu, snaˇzící se pˇrekonat problém ˇskálovatelnosti, kter´y tvoˇrí zásadní omezení pˇri evolučním návrhu. Dosud bylo v oblasti developmentu uvedeno mnoho model˚u a technik, včetně jejich aplikací v r˚uzn´ych oborech. Tato doktorská práce pˇredstavuje novou tˇrídu model˚u pro development, nazvanou development zaloˇzen´y na instrukcích. Základním rysem tohoto pˇrístupu je evoluce aplikačně-specifick´ych program˚u, sestávajících z jednoduch´ych instrukcí, podobně jako v genetickém programování vyuˇzívajícím lineární reprezentaci. Koncept program˚u ve své podstatě umoˇzňuje realizovat univerzální v´ypočetní model v závislosti na volbě instrukčního souboru, interpretaci a zp˚usobu vykonávání instrucí. Program v podobě posloupnosti instrukcí tedy umoˇzňuje specifikovat libovoln´y algoritmus, kter´y je v oblasti developmentu chápán jako pˇredpis pro v´yvoj (konstrukci) cílového objektu. Smyslem této práce je aplikace developmentu zaloˇzeného na instrukcích v návrhu generick´ych struktur. Jako vhodná oblast pro úspěˇsnou demonstraci tohoto záměru byly zvoleny kombinační logické obvody. Jsou zavedeny dva odliˇsné pˇrístupy aplikace developmentu zaloˇzeného na instrukcích. První

Abstract. This paper introduces a novel multi-cellular developmental system where cells are placed in a continuous space. Cells communicate by diffusing and perceiving substances in the environment and are able to migrate around following affinities with substance gradients. The optimization process is performed using Echo State neural networks on the problem of minimizing tile size variations in the context of a tiling problem. Experimental results show that problem complexity only impacts the number of substances used, rather than the number of cells, which implies some sort of scalability with regards to the size of the phenotype. Symmetry breaking and robustness are addressed by adding noise as an intrinsic property of the model. A (positive) side effect is that the resulting model produces very robust solutions with efficient self-healing behavior in the presence of perturbations never met before. 1