Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?
Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur Graph Search.
Concept
Tierra (simulation informatique)
Résumé
Tierra is a computer simulation developed by ecologist Thomas S. Ray in the early 1990s in which computer programs compete for time (central processing unit (CPU) time) and space (access to main memory). In this context, the computer programs in Tierra are considered to be evolvable and can mutate, self-replicate and recombine. Tierra's virtual machine is written in C. It operates on a custom instruction set designed to facilitate code changes and reordering, including features such as jump to template (as opposed to the relative or absolute jumps common to most instruction sets).
The basic Tierra model has been used to experimentally explore in silico the basic processes of evolutionary and ecological dynamics. Processes such as the dynamics of punctuated equilibrium, host-parasite co-evolution and density-dependent natural selection are amenable to investigation within the Tierra framework. A notable difference between Tierra and more conventional models of evolutionary computation, such as genetic algorithms, is that there is no explicit, or exogenous fitness function built into the model. Often in such models there is the notion of a function being "optimized"; in the case of Tierra, the fitness function is endogenous: there is simply survival and death.
According to Thomas S. Ray and others, this may allow for more "open-ended" evolution, in which the dynamics of the feedback between evolutionary and ecological processes can itself change over time (see evolvability), although this claim has not been realized – like other digital evolution systems, it eventually reaches a point where novelty ceases to be created, and the system at large begins either looping or ceases to 'evolve'. The issue of how true open-ended evolution can be implemented in an artificial system is still an open question in the field of artificial life.
Mark Bedau and Norman Packard developed a statistical method of classifying evolutionary systems and in 1997, Bedau et al.
Source officielle
À propos de ce résultat
Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.