high performance computing on graphics processing units: hgpu.org

The model to be described is based on the findings of evolutionary developmental biology (evo devo) and is determined to emphasize the fundamental importance of the development process additionally to genetic material, forming virtual embryos and furthermore Artificial Neural Networks (ANN). It was originally developed by Ronald Thenius in 2008 and extended by Michael Bodi 2009 [1]. Several Genes are forming toolkits, turning on and off genes that are needed for certain adaptive reactions to the development surrounding leading to diverse shapes and forms emerging from the same genetic base [2]. The aim of the model is amongst others to create certain patterns of ANN which match Neural Networks found in real embryos. This is to be achieved by provided respectively cell emitted substances which act as morphogens and hence form the embryo and the resulting neural network by triggering certain cellular behaviors. The concept used in this model is based on multi-agent simulations where each cell acts and reacts individually to any influences applied from its surroundings. These influences may consist of the existence of specific morphogens near the cell or the presence of other cells. Potential reactions of the cell on these influences are proliferation, development to a node within the ANN or death. Apart from that a cell can emit chemical morphogens itself and therefore trigger these behaviors or the emittance of further substances within other cells in its vicinity. This process of specialization is considered to be reversible. All possible actions of the cell are defined in the genome contained by each cell. Following the concepts of evo devo an individual embryo is then exposed to evolution altering its genome. Each of the resulting individuals is then measured up against certain fitness ideals and hence allowed to repoduce or not depending on its score. Genomes are considered to be diploid, therefore the reproduction process takes two individuals to form a new one, merging their two genomes into a genuine novel genome during a simplified crossover process. Application of this virtual evolution process already results in self organized behaviour of the embryo cells that form shapes and networks which can be compared to findings in real embryos.