Genetic Algorithms and Artificial Life

Mitchell, Melanie; Forrest, Stephanie

doi:10.1162/artl.1994.1.3.267

Cited by 108 publications

(63 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It has been realized by means of neural networks [62,130], various forms of evolutionary algorithms [112,113], and reinforcement learning [147], in which agents learn from the consequences of their actions through rewards. Some applications of learning include path finding problems [80], multi-agent systems [101], and robotics [31].…”

Section: Dissertation Contributionsmentioning

confidence: 99%

Novel Methods for Learning and Adaptation in Chemical Reaction Networks

Banda¹

2000

View full text Add to dashboard Cite

State-of-the-art biochemical systems for medical applications and chemical computing are application-specific and cannot be re-programmed or trained once fabricated. The implementation of adaptive biochemical systems that would offer flexibility through programmability and autonomous adaptation faces major challenges because of the large number of required chemical species as well as the timing-sensitive feedback loops required for learning. Currently, biochemistry lacks a systems vision on how the userlevel programming interface and abstraction with a subsequent translation to chemistry should look like. By developing adaptation in chemistry, we could replace multiple hardwired systems with a single programmable template that can be (re)trained to match a desired input-output profile benefiting smart drug delivery, pattern recognition, and chemical computing.I aimed to address these challenges by proposing several approaches to learning and adaptation in Chemical Reaction Networks (CRNs), a type of simulated chemistry, where species are unstructured, i.e., they are identified by symbols rather than molecular structure, and their dynamics or concentration evolution are driven by reactions and reaction rates that follow mass-action and Michaelis-Menten kinetics.Several CRN and experimental DNA-based models of neural networks exist. However, these models successfully implement only the forward-pass, i.e., the input-weight integration part of a perceptron model. Learning is delegated to a non-chemical system that i computes the weights before converting them to molecular concentrations. Autonomous learning, i.e., learning implemented fully inside chemistry has been absent from both theoretical and experimental research.The research in this thesis offers the first constructive evidence that learning in CRNs is, in fact, possible. I have introduced the original concept of a chemical binary perceptron that can learn all 14 linearly-separable logic functions and is robust to the perturbation of rate constants. That shows learning is universal and substrate-free. To simplify the model I later proposed and applied the "asymmetric" chemical arithmetic providing a compact solution for representing negative numbers in chemistry.To tackle more difficult tasks and to serve more complicated biochemical applications, I introduced several key modular building blocks, each addressing certain aspects of chemical information processing and learning. These parts organically combined into gradually more complex systems. First, instead of simple static Boolean functions, I tackled analog time-series learning and signal processing by modeling an analog chemical perceptron. To store past input concentrations as a sliding window I implemented a chemical delay line, which feeds the values to the underlying chemical perceptron. That allows the system to learn, e.g., the linear moving-average and to some degree predict a highly nonlinear NARMA benchmark series.Another important contribution to the area of chemical learning, which I ...

show abstract

Section: Dissertation Contributionsmentioning

confidence: 99%

Novel Methods for Learning and Adaptation in Chemical Reaction Networks

Banda¹

2000

View full text Add to dashboard Cite

show abstract

“…Hill climbing is likely to find a local optimal value, and to find the true, global optimal solution requires the crossing of a local minimum point, or fitness valley, on the fitness landscape [5]. Genetic algorithms, therefore, use a range of small hill-climbing mutations, along with larger mutations, immigration and complex recombination to locate these global optimum points [6].…”

Section: Introductionmentioning

confidence: 99%

Crossing fitness valleys during the evolution of limpet homing behaviour

Stafford

2010

Open Life Sciences

View full text Add to dashboard Cite

Abstract:Evolution is often considered a gradual hill-climbing process, slowly increasing the fitness of organisms. Here I investigate evolution of homing behaviour in simulated intertidal limpets. While the simulation of homing is only a possible mechanism by which homing may have evolved, the process allows an investigation of how evolution may occur over different fitness landscapes. With some fitness landscapes, in order to evolve path integration as a homing mechanism, a temporary reduction in an organism's fitness was required -since high developmental costs occurred before successful homing strategies evolved. Simple hill-climbing algorithms, therefore, only rarely resulted in the evolution of a functional homing behaviour. The inclusion of trail-following greatly increases the frequency of success of evolution of a path integration strategy. Initially an emergent homing behaviour is formed combining path integration with trail-following. This also demonstrates evolution through exaptation, since in the simulation, the original role of trail-following is likely to be unrelated to homing. Analysis of the fitness landscapes of homing in the presence of trail-following behaviour shows a high variability of fitness, which results in the formation of 'stepping-stones' of high fitness across fitness valleys. By using these steppingstones, simple hill-climbing algorithms can reach the global maximum fitness value.© Versita Sp. z o.o.

show abstract

“…As stressed by Mitchell and Forrest [7] "evolution of artificial systems is an important component of artificial life, providing an important modeling tool." On the contrary, in most CA models all cells obey the same local interaction rule that remains fixed over time.…”

Section: Introductionmentioning

confidence: 99%

Cellular Automata in an Artificial Life perspective

Calabretta¹

1998

Cellular Automata: Research Towards Industry

View full text Add to dashboard Cite

Genetic Algorithms and Artificial Life

Cited by 108 publications

References 29 publications

Novel Methods for Learning and Adaptation in Chemical Reaction Networks

Novel Methods for Learning and Adaptation in Chemical Reaction Networks

Crossing fitness valleys during the evolution of limpet homing behaviour

Cellular Automata in an Artificial Life perspective

Contact Info

Product

Resources

About