Adding Learning to the Cellular Development of Neural Networks: Evolution and the Baldwin Effect

1993 ◽  
Vol 1 (3) ◽  
pp. 213-233 ◽  
Author(s):  
Frédéric Gruau ◽  
Darrell Whitley
Keyword(s):  
Neural Networks ◽  
Boolean Functions ◽  
Development Process ◽  
Developmental Process ◽  
Fitness Landscape ◽  
Biological Cell ◽  
Baldwin Effect ◽  
Genetic Search ◽  
Lamarckian Evolution ◽  
The Baldwin Effect

A grammar tree is used to encode a cellular developmental process that can generate whole families of Boolean neural networks for computing parity and symmetry. The development process resembles biological cell division. A genetic algorithm is used to find a grammar tree that yields both architecture and weights specifying a particular neural network for solving specific Boolean functions. The current study particularly focuses on the addition of learning to the development process and the evolution of grammar trees. Three ways of adding learning to the development process are explored. Two of these exploit the Baldwin effect by changing the fitness landscape without using Lamarckian evolution. The third strategy is Lamarckian in nature. Results for these three modes of combining learning with genetic search are compared against genetic search without learning. Our results suggest that merely using learning to change the fitness landscape can be as effective as Lamarckian strategies at improving search.

scholarly journals Development and the Baldwin Effect

2004 ◽  
Vol 10 (1) ◽  
pp. 39-63 ◽  
Author(s):  
Keith L. Downing
Keyword(s):  
Strong Correlation ◽  
Behavioral Plasticity ◽  
Developmental Process ◽  
Fitness Landscape ◽  
Natural World ◽  
Baldwin Effect ◽  
Developmental Mechanism ◽  
Phenotype Space ◽  
The Baldwin Effect ◽  
Proper Balance

Baldwin's classic hypothesis states that behavioral plasticity can speed evolution by (a) smoothing the fitness landscape and (b) indirect genetic assimilation of acquired characteristics. This latter phase demands a strong correlation between genotype and phenotype space. But the natural world shows signs of this correlation at only a very coarse level, since the intervening developmental process greatly complicates the mapping from genetics to physiology and ethology. Hence, development appears to preclude a strong Baldwin effect. However, by adding a simple developmental mechanism to Hinton and Nowlan's classic model of the Baldwin effect, and by allowing evolution to determine the proper balance between direct and indirect mapping of genome to phenotype, this research reveals several different effects of development on the Baldwin effect, some promoting and others inhibiting. Perhaps the most interesting result is an evolved cooperation between direct blueprints and indirect developmental recipes in searching for unstructured and partially structured target patterns in large, needle-in-the-haystack fitness landscapes.

scholarly journals How to Shift Bias: Lessons from the Baldwin Effect

1996 ◽  
Vol 4 (3) ◽  
pp. 271-295 ◽  
Author(s):  
Peter Turney
Keyword(s):  
Learning Algorithm ◽  
Inductive Learning ◽  
Learning Algorithms ◽  
Darwinian Evolution ◽  
Baldwin Effect ◽  
Lamarckian Evolution ◽  
Inheritance Of Acquired Characteristics ◽  
The Baldwin Effect ◽  
Fixed Bias ◽  
Shift Of Bias

An inductive learning algorithm takes a set of data as input and generates a hypothesis as output. A set of data is typically consistent with an infinite number of hypotheses; therefore, there must be factors other than the data that determine the output of the learning algorithm. In machine learning, these other factors are called the bias of the learner. Classical learning algorithms have a fixed bias, implicit in their design. Recently developed learning algorithms dynamically adjust their bias as they search for a hypothesis. Algorithms that shift bias in this manner are not as well understood as classical algorithms. In this paper, we show that the Baldwin effect has implications for the design and analysis of bias shifting algorithms. The Baldwin effect was proposed in 1896 to explain how phenomena that might appear to require Lamarckian evolution (inheritance of acquired characteristics) can arise from purely Darwinian evolution. Hinton and Nowlan presented a computational model of the Baldwin effect in 1987. We explore a variation on their model, which we constructed explicitly to illustrate the lessons that the Baldwin effect has for research in bias shifting algorithms. The main lesson is that it appears that a good strategy for shift of bias in a learning algorithm is to begin with a weak bias and gradually shift to a strong bias.

scholarly journals The Influence of Learning on Evolution: A Mathematical Framework

2009 ◽  
Vol 15 (2) ◽  
pp. 227-245 ◽  
Author(s):  
Ingo Paenke ◽  
Tadeusz J. Kawecki ◽  
Bernhard Sendhoff
Keyword(s):  
Fitness Landscape ◽  
Simulation Models ◽  
Mathematical Framework ◽  
Gain Function ◽  
Baldwin Effect ◽  
Positive Gain ◽  
Proportional Change ◽  
Acceleration And Deceleration ◽  
The Baldwin Effect ◽  
Evolutionary Fitness

The Baldwin effect can be observed if phenotypic learning influences the evolutionary fitness of individuals, which can in turn accelerate or decelerate evolutionary change. Evidence for both learning-induced acceleration and deceleration can be found in the literature. Although the results for both outcomes were supported by specific mathematical or simulation models, no general predictions have been achieved so far. Here we propose a general framework to predict whether evolution benefits from learning or not. It is formulated in terms of the gain function, which quantifies the proportional change of fitness due to learning depending on the genotype value. With an inductive proof we show that a positive gain-function derivative implies that learning accelerates evolution, and a negative one implies deceleration under the condition that the population is distributed on a monotonic part of the fitness landscape. We show that the gain-function framework explains the results of several specific simulation models. We also use the gain-function framework to shed some light on the results of a recent biological experiment with fruit flies.

Evolving Neural Networks Using the “Baldwin Effect”

1995 ◽  
pp. 333-336 ◽  
Author(s):  
Egbert J. W. Boers ◽  
Marko V. Borst ◽  
Ida G. Sprinkhuizen-Kuyper
Keyword(s):  
Neural Networks ◽  
Baldwin Effect ◽  
The Baldwin Effect

On the Evolution of Multicellularity and Eusociality

1999 ◽  
Vol 5 (1) ◽  
pp. 1-15 ◽  
Author(s):  
Larry Bull
Keyword(s):  
Fitness Landscape ◽  
Baldwin Effect ◽  
Daughter Cells ◽  
Unicellular Organisms ◽  
Subsequent Effect ◽  
The Baldwin Effect

In this article versions of the abstract NKC model are used to examine the conditions under which two significant evolutionary phenomena—multicellularity and eusociality—are likely to occur and why. First, comparisons in evolutionary performance are made between simulations of unicellular organisms and very simple multicellular-like organisms, under varying conditions. The results show that such multicellularity without differentiation appears selectively neutral, but that differentiation to soma (nonreproductives) proves beneficial as the amount of epistasis in the fitness landscape increases. This is explained by considering mutations in the generation of daughter cells and their subsequent effect on the propagule's fitness. This is interpreted as a simple example of the Baldwin effect. Second, the correspondences between multicellularity and eusociality are highlighted, particularly that both contain individuals who do not reproduce. The same process is then used to explain the emergence of eusocial colonies.

The Dynamic Changes in Roles of Learning through the Baldwin Effect

2007 ◽  
Vol 13 (1) ◽  
pp. 31-43 ◽  
Author(s):  
Reiji Suzuki ◽  
Takaya Arita
Keyword(s):  
Phenotypic Plasticity ◽  
Fitness Landscape ◽  
Evolutionary Model ◽  
Hill Climbing ◽  
Complex Environments ◽  
Baldwin Effect ◽  
Epistatic Interactions ◽  
Genetic Robustness ◽  
The Baldwin Effect ◽  
Insight Into

The interaction between evolution and learning called the Baldwin effect is a two-step evolutionary scenario caused by the balances between benefit and cost of learning in general. However, little is known about the dynamic evolution of these balances in complex environments. Our purpose is to give a new insight into the benefit and cost of learning by focusing on the quantitative evolution of phenotypic plasticity under the assumption of epistatic interactions. For this purpose, we have constructed an evolutionary model of quantitative traits by using an extended version of Kauffman's NK fitness landscape. Phenotypic plasticity is introduced into our model; whether each phenotype is plastic or not is genetically defined, and plastic phenotypes can be adjusted by learning. The simulation results clearly show that drastic changes in roles of learning cause three-step evolution through the Baldwin effect and also cause the evolution of genetic robustness against mutations. We also conceptualize four different roles of learning by using a hill-climbing image of a population on a fitness landscape.

On the Baldwin Effect

1999 ◽  
Vol 5 (3) ◽  
pp. 241-246 ◽  
Author(s):  
Larry Bull
Keyword(s):  
Fitness Landscape ◽  
Higher Learning ◽  
Learning Rate ◽  
Adaptation Process ◽  
High Learning ◽  
Baldwin Effect ◽  
Nk Model ◽  
Learning Rates ◽  
Rate Of Learning ◽  
The Baldwin Effect

In this article the effects of altering the rate and amount of learning on the Baldwin effect are examined. Using a version of the abstract tunable NK model, it is shown that the adaptation process is sensitive to the rate of learning, particularly as the correlation of the underlying fitness landscape varies. Typically a high learning rate proves most beneficial as landscape correlation decreases. It is also shown that the amount of learning can have a significant effect on the adaptation process, where increased amounts of learning prove beneficial under higher learning rates on uncorrelated landscapes.

Sign in / Sign up

Export Citation Format

Share Document