EvoShelf : A System for Managing and Exploring Evolutionary Data Timothy Davison 1 , Sebastian von Mammen 1 , and Christian Jacob 1,2 1 Dept. of Computer Science, Faculty of Science 2 Dept. of Biochemistry & Molecular Biology, Faculty of Medicine University of Calgary, Canada {tbdaviso,s.vonmammen,cjacob}@ucalgary.ca Abstract. Systems that utilize evolutionary computation produce large amounts of data. Quite often, this data has a convenient visual represen- tation. However, managing and visualizing evolutionary data can be a difficult and onerous task. By employing techniques used in photo man- agement software, we have produced a system that helps to organize and visualize evolutionary data while retaining a complete record of a simulation. By means of a simple plugin architecture this system can be extended to import data produced by arbitrary evolutionary systems. We present the system’s architecture, its features, and we provide a com- prehensive example, highlighting its advantages in applied research. 1 Introduction Evolutionary systems produce large amounts of data. Beyond the obvious data (such as the genotype and phenotype of an individual), there is a considerable amount of meta-data produced as well. Such data includes the hereditary data, fitness values, and other attributes of the evolutionary computation approach being employed. It is common to manage experimental data by means of a file-system browser, such as the Finder in Mac OS X, and Windows Explorer in Microsoft Windows. Searching or organizing individuals according to various criteria is a laborious task in such systems. Consider a system that organizes the individuals produced by an experiment into sub-directories by generation, giving each individual its own file containing its genotype, and phenotype, along with meta-data such as fitness, or genealogy. Filtering these individuals by fitness value would be a difficult task with either file-system browser. An evolutionary system may employ an interface of its own for browsing the data that it produces. In this case, the visualization procedures and the manage- ment of the genotype/phenotype data are typically implemented specifically for the one evolutionary system. However, the universality of evolutionary algorith- mic approaches renders generic visualization and data management techniques valuable across various application domains. In a way, the situation is very similar to managing individual (digitized) image and music collections. Such libraries can easily consist of thousands of items. A R. Schaefer et al. (Eds.): PPSN XI, Part II, LNCS 6239, pp. 310–319, 2010. c  Springer-Verlag Berlin Heidelberg 2010