Tangen, Uwe. “An evolvable micro-controller or what’s new about mutations?.” In Proceedings of the 4th Annual Conference on Genetic and Evolutionary Computation , pp. 178-186. 2002.
URL1
Un-supervised learning of complex software projects is still an issue. Additionally, the objective to create open-ended evolutionary systems is lacking a confident realization. Long-term evolutionary behavior results in noisy regimes or stable fixed points including limit cycles. The evolving organizations are more or less simple or directly reflecting the programmer’s point of view of a complex organization. One possible way to over come these limitations is to drastically increase population sizes and look at longer timescales. The work reported serves two purposes: the introduction of an evolving micro-controller inclusive evolving assembler code on the new massively parallel configurable hardware MereGen™ and it provides new insights on how mutation probabilities may influence the evolutionary outcome of evolving populations.
With 648 MBytes of fast SRAM and about a Gigabyte of SDRAM, millions of programs and thousands of micro-controllers can be evolved concurrently in hardware. Typical speedups are about five orders of magnitude compared with current day high-end PCs. Thus, longterm evolution of large populations is no longer the bottleneck and limi tations in our understanding and the observability of evolving systems immediately grabs the focus of our attention in evolutionary experiments.