Robustness, evolvability, and optimality of evolutionary neural networks

In a typical optimization problem, the main goal is to search for the appropriate values of the variables that provide the optimal solution of the given function. In artificial neural networks (ANN), this translates to the minimization of the error surface during training such that misclassification is minimized during generalization. However, since optimal training performance does not necessarily imply optimal generalization due to the possibility of overfitting or underfitting, we developed SEPA (Structure Evolution and Parameter Adaptation) which addressed these issues by simultaneously evolving ANN structure and weights. Since SEPA primarily relies on the perturbation function to bring variation in its population, this follow-up study aims to find out SEPAs evolvability, optimality, and robustness in other perturbation functions. Our findings indicate that SEPAs optimal generalization performances are stable and robust from the effect of the different perturbation functions. This is due to the feedback loop between its architecture evolution and weight adaptation such that any shortcoming of the former is compensated by the latter, and vice versa. Our results strongly suggest that proper ANN design requires simultaneous adaptation of ANN structure and weights to avoid one-sided or bias convergence to either the weight or architecture space.     

Evo-devo: extending the evolutionary synthesis

Evolutionary developmental biology (evo-devo) explores the mechanistic relationships between the processes of individual development and phenotypic change during evolution. Although evo-devo is widely acknowledged to be revolutionizing our understanding of how the development of organisms has evolved, its substantial implications for the theoretical basis of evolution are often overlooked. This essay identifies major theoretical themes of current evo-devo research and highlights how its results take evolutionary theory beyond the boundaries of the Modern Synthesis.     

Neural mechanisms underlying the evolvability of behaviour

The complexity of nervous systems alters the evolvability of behaviour. Complex nervous systems are phylogenetically constrained; nevertheless particular species-specific behaviours have repeatedly evolved, suggesting a predisposition towards those behaviours. Independently evolved behaviours in animals that share a common neural architecture are generally produced by homologous neural structures, homologous neural pathways and even in the case of some invertebrates, homologous identified neurons. Such parallel evolution has been documented in the chromatic sensitivity of visual systems, motor behaviours and complex social behaviours such as pair-bonding. The appearance of homoplasious behaviours produced by homologous neural substrates suggests that there might be features of these nervous systems that favoured the repeated evolution of particular behaviours. Neuromodulation may be one such feature because it allows anatomically defined neural circuitry to be re-purposed. The developmental, genetic and physiological mechanisms that contribute to nervous system complexity may also bias the evolution of behaviour, thereby affecting the evolvability of species-specific behaviour.     

Predicting evolutionary potential: A numerical test of evolvability measures

Despite sophisticated mathematical models, the theory of microevolution is mostly treated as a qualitative rather than a quantitative tool. Numerical measures of selection, constraints, and evolutionary potential are often too loosely connected to theory to provide operational predictions of the response to selection. In this paper, we study the ability of a set of operational measures of evolvability and constraint to predict short‐term selection responses generated by individual‐based simulations. We focus on the effects of selective constraints under which the response in one trait is impeded by stabilizing selection on other traits. The conditional evolvability is a measure of evolutionary potential explicitly developed for this situation. We show that the conditional evolvability successfully predicts rates of evolution in an equilibrium situation, and further that these equilibria are reached with characteristic times that are inversely proportional to the fitness load generated by the constraining characters. Overall, we find that evolvabilities and conditional evolvabilities bracket responses to selection, and that they together can be used to quantify evolutionary potential on time scales where the G‐matrix remains relatively constant.     

Evolution: the long evolutionary reach of viruses

The structure of a phage capsid protein provides good evidence this phage shares ancestry with an animal virus. In this and similar cases, either the viral lineages predate the emergence of the three contemporary domains of life, or viruses have been leaping the phylogenetic chasms that separate the domains.     

The evolvability of herkogamy: Quantifying the evolutionary potential of a composite trait

Accurate estimates of trait evolvabilities are central to predicting the short‐term evolutionary potential of populations, and hence their ability to adapt to changing environments. We quantify and evaluate the evolvability of herkogamy, the spatial separation of male and female structures in flowers, a key floral trait associated with variation in mating systems. We compiled genetic‐variance estimates for herkogamy and related floral traits, computed evolvabilities, and compared these among trait groups and among species differing in their mating systems. When measured in percentage of its own size, the median evolvability of herkogamy was an order of magnitude greater than the evolvability of other floral size measurements, and was generally not strongly constrained by genetic covariance between its components (pistil and stamen lengths). Median evolvabilities were similar across mating systems, with only a tendency toward reduction in highly selfing taxa. We conclude that herkogamy has the potential to evolve rapidly in response to changing environments. This suggests that the extensive variation in herkogamy commonly observed among closely related populations and species may result from rapid adaptive tracking of fitness optima determined by variation in pollinator communities or other selective factors.     

Rupert Riedl and the re-synthesis of evolutionary and developmental biology: body plans and evolvability

This paper reviews the scientific career of Rupert Riedl and his contributions to evolutionary biology. Rupert Riedl, a native of Vienna, Austria, began his career as a marine biologist who made important contributions to the systematics and anatomy of major invertebrate groups, as well as to marine ecology. When he assumed a professorship at the University of North Carolina in 1968, the predominant thinking in evolutionary biology focused on population genetics, to the virtual exclusion of most of the rest of biology. In this atmosphere Riedl developed his "systems theory" of evolution, which emphasizes the role of functional and developmental integration in limiting and enabling adaptive evolution by natural selection. The main objective of this theory is to account for the observed patterns of morphological evolution, such as the conservation of body plans. In contrast to other "alternative" theories of evolution, Riedl never denied the importance of natural selection as the driving force of evolution, but thought it necessary to contextualize natural selection with the organismal boundary conditions of adaptation. In Riedl's view development is the most important factor besides natural selection in shaping the pattern and processes of morphological evolution.     

The evolutionary dynamics of evolvability in a gene network model

Evolvability, the ability of populations to adapt, has recently emerged as a major unifying concept in biology. Although the study of evolvability offers new insights into many important biological questions, the conceptual bases of evolvability, and the mechanisms of its evolution, remain controversial. We used simulated evolution of a model of gene network dynamics to test the contentious hypothesis that natural selection can favour high evolvability, in particular in sexual populations. Our results conclusively demonstrate that fluctuating natural selection can increase the capacity of model gene networks to adapt to new environments. Detailed studies of the evolutionary dynamics of these networks establish a broad range of validity for this result and quantify the evolutionary forces responsible for changes in evolvability. Analysis of the genotype–phenotype map of these networks also reveals mechanisms connecting evolvability, genetic architecture and robustness. Our results suggest that the evolution of evolvability can have a pervasive influence on many aspects of organisms.     

Robustness, evolvability, and neutrality

Biological systems, from macromolecules to whole organisms, are robust if they continue to function, survive, or reproduce when faced with mutations, environmental change, and internal noise. I focus here on biological systems that are robust to mutations and ask whether such systems are more or less evolvable, in the sense that they can acquire novel properties. The more robust a system is, the more mutations in it are neutral, that is, without phenotypic effect. I argue here that such neutral change--and thus robustness--can be a key to future evolutionary innovation, if one accepts that neutrality is not an essential feature of a mutation. That is, a once neutral mutation may cause phenotypic effects in a changed environment or genetic background. I argue that most, if not all, neutral mutations are of this sort, and that the essentialist notion of neutrality should be abandoned. This perspective reconciles two opposing views on the forces dominating organismal evolution, natural selection and random drift: neutral mutations occur and are especially abundant in robust systems, but they do not remain neutral indefinitely, and eventually become visible to natural selection, where some of them lead to evolutionary innovations.     

Evolution: the evolvability enigma

A report that a switch of a yeast protein to a 'prion' state triggers diverse phenotypic changes has prompted re-examination of the processes of evolution. To what extent should processes of gene expression and control be interpreted in terms of their capacity to allow future evolution as well as present adaptation?     

Precopulatory behaviour and the evolutionary relationships of Discoglossidae

The phylogenetic relationships among primitive anurans remain controversial, even when several morphological and molecular approaches have been used in an attempt to resolve them. Nevertheless, very little effort has been made to study these relationships with behavioural data. We studied the relation between some precopulatory behaviour characters (male call characteristics and call recognition by females) and the different proposed phylogenies among the anuran genera Alytes, Discoglossus and Bombina . Different acoustic variables were analysed from male calls of different species of these genera; Alytes cisternasii females were used in recognition tests with calls from those species. A phenetic tree generated from male call characteristics grouped Bombina with Discoglossu s and separated these two genera from Alytes, although the tree generated from female response data did not show such a clear result. A. cisternasii females are mostly using some features that are absent in Alytes calls in order to recognize male calls. The two characters, mapped onto the three alternative phylogenetic proposals, did not resolve the trichotomy among the three genera, but suggested that the studied behavioural characters have a clear phylogenetic load.     

Success-biased social learning: Cultural and evolutionary dynamics

Success bias is a social learning strategy whereby learners tend to acquire the cultural variants of successful individuals. I develop a general model of success-biased social learning for discrete cultural traits with stochastic payoffs, and investigate its dynamics when only two variants are present. I find that success bias inherently favors rare variants, and consequently performs worse than unbiased imitation (i.e. random copying) when success payoffs are at least mildly stochastic and the optimal variant is common. Because of this weakness, success bias fails to replace unbiased imitation in an evolutionary model when selection is fairly weak or when the environment is relatively stable, and sometimes fails to invade at all. I briefly discuss the optimal strength of success bias, the complicated nature of defining success in social learning contexts, and the value of variant frequency as an important source of information to social learners. I conclude with predictions regarding the prevalence of success bias in different behavioral domains.     

Pleiotropic mutation, modularity and evolvability

The relationship between pleiotropy and the rate of evolution of a phenotypic character (evolvability) in a population is explored using computer simulations. I present results that suggest the rate of evolution of a phenotypic character may not decline when that character is pleiotropically associated to an increasing number of other characters, provided that the characters are under pure directional selection such that they are far from their optima relative to the average magnitude of a mutation. These conditions may be relevant during adaptive radiations. Adding pleiotropic associations to a set of characters in which one is under directional selection and the other is under stabilizing selection increases the rate of adaptation of the character under directional selection provided that the new characters that come to be pleiotropically associated are under directional selection. Thus, increasing the number of pleiotropic associations under these conditions increases the rate of adaptation of a character.     

Some relations between stereotyped suckling in piglets and exploratory behaviour and discrimination reversal learning in adult swine

The relationship of stereotyped suckling positions in artificially fed and sow-fed infant swine to the behaviour of adult animals was investigated. Suckling behaviour was observed, and four animals from each feeding condition exhibiting low suckling stereotypy and four from each feeding condition exhibiting high suckling stereotypy were selected. Animals were tested at five months of age in exploratory situations and on a two-window position discrimination and reversals.Low suckling stereotypy and artificially reared animals made fewer errors on the first discrimination reversals than high stereotypy and sow-fed animals. They were also less emotional on the first reversals. Over nine reversals artificially reared animals made significantly fewer errors than sow-fed subjects and low stereotype sow-fed animals made fewer errors than high stereotypy sow-reared animals. Low stereotypy subjects explored more objects and over a longer period of time than high stereotypy subjects. It was concluded that stereotyped suckling in piglets is related to behaviour exhibited by the adult in problem-solving situations.     

Quantifying the exploratory behaviour of Amphibalanus amphitrite cyprids

The behavioural response of cypris larvae from A. amphitrite (=Balanus amphitrite) exploring three model glass surfaces is quantified by close-range microscopy. Step length and step duration measurements reveal a response to both surface properties and flow. Without flow, 2-day-old cyprids took larger steps with shorter step duration on hydrophilic glass surfaces (bare and NH₂-treated) vs hydrophobic glass (CH₃-treated). These parameters suggest a more detailed, local inspection of hydrophobic surfaces and a more extensive exploration for hydrophilic surfaces. Cyprids under flow took longer steps and exhibited shorter probing times on hydrophobic glass. On hydrophilic glass, cyprids increased their step duration under flow. This active response is attributed to drag and lift forces challenging the cyprids' temporary anchoring to the substratum. Seven-day-old cyprids showed almost no discrimination between the model surfaces. Microscopic-scale observation of cyprid exploration is expected to provide new insights into interactions between cyprids and surfaces.