This paper focuses on the phenomena of evolution whose appearance is notable because no explicit mutation, recombination or artificial selection operators are introduced. We call the system self-evolving because every variation is performed by the objects themselves in their machine form. Keywords: artificial chemistry, autocatalytic reaction system, molecular computing, prebiotic evolution, self-organization, self-programming 1

Adaptation of ecological systems to their environments is commonly viewed through some explicit fitness function defined a priori by the experimenter, or measured a posteriori by estimations based on population size and/or reproductive rates. These methods do not capture the role of environmental complexity in shaping the selective pressures that control the adaptive process. Ecological simulations enabled by computational tools such as the Latent Energy Environments (LEE) model allow us to characterize more closely the effects of environmental complexity on the evolution of adaptive behaviors. LEE is described in this paper. Its motivation arises from the need to vary complexity in controlled and predictable ways, without assuming the relationship of these changes to the adaptive behaviors they engender. This goal is achieved through a careful characterization of environments in which different forms of "energy" are well-defined. A genetic algorithm using endogenous fitness and local ...

This article reviews the growing body of scientific work in Artificial Chemistry. First, common motivations and fundamental concepts are introduced. Second, current research activities are discussed along three application dimensions: modelling, information processing and optimization. Finally, common phenomena among the different systems are summarized. It is argued here that Artificial Chemistries are "the right stuff" for the study of pre-biotic and bio-chemical evolution, and they provide a productive framework for questions regarding the origin and evolution of organizations in general. Furthermore, Artificial Chemistries have a broad application range to practical problems as shown in this review.

We discuss a system of autocatalytic sequences of binary numbers. Sequences come in two forms, a 1dimensional form (operands) and a 2-dimensional form (operators) that are able to react with each other. The resulting reaction network shows signs of emerging metabolisms. We discuss the general framework and examine specific interactions for a system with strings of length 4 bits. A selfmaintaining network of string types and parasitic interactions are shown to exist. Introduction Published in: Proceedings ARTIFICIAL LIFE IV R. Brooks and P. Maes (Eds.) MIT Press, Cambridge, MA, 1994 pp. 109 --- 118 Sequences of binary numbers are the most primitive form of information storage we know today. They are able to code any kind of man-made information, be it still or moving images, sound waves and other sensory stimulations, be it written language or the rules of mathematics, just to name a few. As the success of vonNeumann computers has shown over the last 50 years, binary sequences are also...

Formalized rules for protein-protein interactions have recently been introduced to represent the binding and enzymatic activities of proteins in cellular signaling. Rules encode an understanding of how a system works in terms of the biomolecules in the system and their possible states and interactions. A set of rules can be as easy to read as a diagrammatic interaction map, but unlike most such maps, rules have precise interpretations. Rules can be processed to automatically generate a mathematical or computational model for a system, which enables explanatory and predictive insights into the system’s behavior. Rules are independent units of a model specification that facilitate model revision. Instead of changing a large number of equations or lines of code, as may be required in the case of a conventional mathematical model, a protein interaction can be introduced or modified simply by adding or changing a single rule that represents the interaction of interest. Rules can be defined and visualized by using graphs, so no specialized training in mathematics or computer science is necessary to create models or to take advantage of the representational precision of rules. Rules can be encoded in a machine-readable format to enable electronic storage and exchange of models, as well as basic knowledge about protein-protein interactions. Here, we review the motivation for rule-based modeling; applications of the approach; and issues that arise in model specification, simulation, and testing. We also discuss rule visualization and exchange and the software available for rule-based modeling.

This paper demonstrates a new method of programming artificial chemistries. It uses the emerging capabilities of the system's dynamics for information processing purposes. By evolution of metabolisms that act as control programs for a small robot one achieves the adaptation of the internal metabolic pathways as well as the selection of the most relevant available exteroreceptors. The underlying artificial chemistry evolves efficient information processing pathways with most benefit for the desired task, robot navigation. The results show certain relations to biological systems like motile bacteria.

An artificial life entertainment-software product called Creatures was released in Europe in late 1996 and in the United States and Japan in mid-1997. When installed on a domestic computer (PC or Macintosh), each Creatures CD-ROM creates a virtual world in which autonomous software agents exist. The agents, known as "norns", interact with the human user, with each other, and with objects in their virtual world. Each norn coordinates perception and action via its own modular recurrent neural network: Each network has Hebbian learning, plus diffuse modulation of activity via a "hormonal" system that is part of that norn's "biochemistry". Details of each norn's neural network and biochemistry are genetically specified, and norns can breed via sexual reproduction. In the reproduction process, genetic material may be mutated and may also be subjected to "gene duplications" that enable potentially unlimited increases in complexity of the norns' design. Over 500,000 Creatures CD-ROMS have now been sold. As each installed copy of Creatures can support 5 to 10 simultaneously existing individual norns, it seems reasonable to estimate that there are up to 5 million norns existing in the "cyberspace" provided by the global Creatures user community.

In this paper we consider evolutionary dynamics of catalytically active species with a distinct genotype -- phenotype relationship. Folding landscapes of RNA-molecules serve as a paradigm for this relationship with essential neutral properties. This landscape itself is partitioned by phenotypes (realized as secondary structures). To each genotype (represented as sequence) a structure is assigned in a unique way. The set of all sequences which map into a particular structure is modeled as random graph in sequence space (the so-called neutral network). A catalytic network is realized as a random digraph with maximal out-degree 2 and secondary structures as vertex set. Studying a population of catalytic RNA-molecules shows significantly different behavior compared to a deterministic description: hypercycles are able to co-exist and survive resp. a parasite with superior catalytic support. A "switching" between different dynamical organizations of the network can be obs...

In an algorithmic artificial chemistry the objects (molecules) are data and the interactions (reactions) among them are defined by an algorithm. The same object can appear in two forms: (1) as a machine (operator) or (2) as data (operand). Thus, the same object can, on the one hand, process other objects or, on the other hand, it can be processed. This dualism enables to implicitly define a constructive artificial chemistry which exhibits quite complex behavior. Remarkably, even evolutionary behavior emerged in our experiments, without defining any explicit variation operators or fitness-function. In addition to microscopic methods (e.g., monitoring the actions of single molecules) and macroscopic measures (e.g., diversity or complexity) we developed a stepwise mesoscopic analysis method based on classification and dynamic clustering. Knowledge about the system is accumulated by an iterative process in which measuring tools (classificators) extract information which i...

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii I Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 A. Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1. The #arti#cial life bridge" . . . . . . . . . . . . . . . . . . . . . . 1 2. From nature to technology . . . . . . . . . . . . . . . . . . . . . 3 3. From technology to nature . . . . . . . . . . . . . . . . . . . . . 4 B. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 II Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 A. Background: Machine learning . . . . . . . . . . . . . . . . . . . . . 10 1. Evolutionary algorithms . . . . . . . . . . . . . . . . . . . . . . . 11 2. Endogenous #tness . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3. Reinforcement learning . . . . . . . . . . . . . . . . . . . . . . . 18 B. Local selection . . . . . . . . . . . . . . . . . . . . . . . . . . ...