Evolving Plasticity for Autonomous Learning under Changing Environmental Conditions.

A fundamental aspect of learning in biological neural networks is the plasticity property which allows them to modify their configurations during their lifetime. Hebbian learning is a biologically plausible mechanism for modeling the plasticity property in artificial neural networks (ANNs), based on the local interactions of neurons. However, the emergence of a coherent global learning behavior from local Hebbian plasticity rules is not very well understood. The goal of this work is to discover interpretable local Hebbian learning rules that can provide autonomous global learning. To achieve this, we use a discrete representation to encode the learning rules in a finite search space. These rules are then used to perform synaptic changes, based on the local interactions of the neurons. We employ genetic algorithms to optimize these rules to allow learning on two separate tasks (a foraging and a prey-predator scenario) in online lifetime learning settings. The resulting evolved rules converged into a set of well-defined interpretable types, that are thoroughly discussed. Notably, the performance of these rules, while adapting the ANNs during the learning tasks, is comparable to that of offline learning methods such as hill climbing.

The many functions of vocal learning.

The capacity to learn novel vocalizations has evolved convergently in a wide range of species. Courtship songs of male birds or whales are often treated as prototypical examples, implying a sexually selected context for the evolution of this ability. However, functions of learned vocalizations in different species are far more diverse than courtship, spanning a range of socio-positive contexts from individual identification, social cohesion, or advertising pair bonds, as well as agonistic contexts such as territorial defence, deceptive alarm calling or luring prey. Here, we survey the diverse usages and proposed functions of learned novel signals, to build a framework for considering the evolution of vocal learning capacities that extends beyond sexual selection. For each function that can be identified for learned signals, we provide examples of species using unlearned signals to accomplish the same goals. We use such comparisons to generate hypotheses concerning when vocal learning is adaptive, given a particular suite of socio-ecological traits. Finally, we identify areas of uncertainty where improved understanding would allow us to better test these hypotheses. Considering the broad range of potential functions of vocal learning will yield a richer appreciation of its evolution than a narrow focus on a few prototypical species. This article is part of the theme issue 'Vocal learning in animals and humans'.

The evolution of sexual imprinting through reinforcement.

Reinforcement is the process whereby assortative mating evolves due to selection against costly hybridization. Sexual imprinting could evolve as a mechanism of reinforcement, decreasing hybridization, or it could potentially increase hybridization in genetically purebred offspring of heterospecific social pairs. We use deterministic population genetic simulations to explore conditions under which sexual imprinting can evolve through reinforcement. We demonstrate that a sexual imprinting component of female preference can evolve as a one-allele assortative mating mechanism by reducing the risk of hybridization, and is generally effective at causing trait divergence. However, imprinting often evolves to be a component rather than the sole determinant of female preference. The evolution of imprinting has the unexpected side effect of homogenizing existing innate preference, because the imprinted preference neutralizes any innate preference. We also find that the weight of the imprinting component may evolve to a lower value when migration and divergent selection are strong and the cost of hybridization is low; these conditions render hybridization adaptive for immigrant females because they can acquire locally adaptive genes by mating with local males. Together, these results suggest that sexual imprinting can itself evolve as part of the speciation process, and in doing so has the capacity to promote or retard divergence through complex interactions.

Entwined engrams: The evolution of associative and non-associative learning.

The nematode Caenorhabditis elegans displays a surprisingly sophisticated behavioral repertoire that includes the utilization of both associative and non-associative forms of learning. Elucidating the molecular basis of learning remains a fundamental, yet daunting, challenge of modern neuroscience. In Pereira and van der Kooy (ref. 2), we described the use of a two input-two output stimuli system to dissociate between associative and non-associative learning and between memory acquisition and retrieval processes. Briefly, one finding indicated that after training with the odorant isoamyl alcohol (IsoA), we could preferentially retrieve either associative or non-associative memory with a choice of either a benzaldehyde (Bnz) or IsoA retrieval stimulus, respectively. Here, we describe how that apparently enigmatic molecular cross wiring of the two forms of memory examined could represent an evolutionary relic of the ancient divergence between non-associative and associative learning. In addition, we extrapolate on the utility and subtleties of using such a system to dissociate and decipher the components of memory in C. elegans.

PHYLOGENETIC ANALYSIS OF LEARNING-RELATED NEUROMODULATION IN MOLLUSCAN MECHANOSENSORY NEURONS.

In spite of significant advances in our understanding of mechanisms of learning and memory in a variety of organisms, little is known about how such mechanisms evolve. Even mechanisms of simple forms of learning, such as habituation and sensitization, have not been studied phylogenetically. Here we begin an evolutionary analysis of learning-related neuromodulation in species related to the well-studied opisthobranch gastropod, Aplysia californica. In Aplysia, increased spike duration and excitability in mechanosensory neurons contribute to several forms of learning-related changes to defensive withdrawal reflexes. The modulatory transmitter serotonin (5-hydroxytryptamine, or 5-HT), is thought to play a critical role in producing these firing property changes. In the present study, we tested mechanosensory homologs of the tail-withdrawal reflex in species related to Aplysia for 5-HT-mediated increases in spike duration and excitability. Criteria used to identify homologous tail-sensory neurons included position, relative size, resting electrical properties, expression of a sensory neuron-specific protein, neuroanatomy, and receptive field. The four ingroup species studied (Aplysia californica, Dolabella auricularia, Bursatella leachii, and Dolabrifera dolabrifera) belong to two clades (two species each) within the family Aplysiidae. In the first clade (Aplysia/Dolabella), we found that the tail-sensory neurons of A. californica and tail-sensory homologs of a closely related species, D. auricularia, responded to bath-applied serotonin in essentially similar fashion: significant increases in spike duration as well as excitability. In the other clade (Dolabrifera/Bursatella), more distantly related to Aplysia, one species (B. leachii) showed spike broadening and increased excitability. However, the other species (D. dolabrifera) showed neither spike broadening nor increased excitability. The firing properties of tail-sensory homologs of D. dolabrifera were insensitive to 5-HT over a wide range of concentrations. We also performed experiments on two outgroup species (Akera bullata and Bulla gouldiana) and found that spike duration was unaffected by 5-HT, whereas excitability was increased. This study suggests that 5-HT-induced spike broadening arose more recently in opisthobranch evolution, whereas 5-HT-induced excitability increase is a more ancestral trait that may have been expressed in the earliest opisthobranchs. Both traits are absent in the aplysiid species D. dolabrifera, demonstrating that a lineage can lose learning-related mechanisms. The phylogenetic variation observed in the present study presents the opportunity to test general models about learning mechanisms and their evolution in unique ways.