Looking For Energy Efficient Genetic Algorithms

When Evolutionary Algorithms (EAs) are applied to optimization problems, two main measures are taken into account to understand their performance: fitness quality and computing time. These two values are used to compare the performance of different versions of an algorithm, different parameter settings of a single algorithm or even compare a particular EA with other available heuristics. Nevertheless, a new trend in computer science tries to contextualize these features under a new perspective: power consumption. This paper presents a preliminary analysis of the standard genetic algorithm, using two well known benchmark problems, considering their fitness quality, the computing time and also the power consumption when battery-powered devices are used to run them. Results show that some of the main parameters of the algorithm have an impact on instantaneous energy consumption -that departs from the expected behavior, and therefore affects the amount of energy required to run the algorithm. Although we are still far from finding a way to design energy-efficient EAs, we think the results open up a new perspective that will enable us to achieve this goal in the future.