The number of seats is limited to 25 due to COVID-19 regulations.
Abstract
Both evolution and learning are known to produce (sometimes spectacular) adaptive solutions. One can rightfully ask whether these processes might share some common features, and whether they can help each other, possibly in the form of one being a "subroutine" in the other and vice versa. Recent models inform us that ecosystem evolution and evolution of genetic regulatory networks (so important in development) can partly be understood as learning processes. Features like Hebbian change in coupling terms, memory capacity, forgetting and graceful degradation all come into play. These investigations are complemented by the proposals that the Bayesian update rule is analogous to the discrete-time replicator equation and that evolving replicator populations can learn about grammatical classes of problems.
The flip side of the coin is how much evolution could happen during cognition. Darwinian neurodynamics proposes that candidate solutions to complex problems, encoded by transient or more permanent neuronal activity patterns, could be subject to bona fide evolutionary dynamics ("the breeding of ideas"). Merits and open questions of this idea will be discussed.
Speaker
Dr. Eörs Szathmáry, MHAS, Professor of Biology.
Insttitute of Evolution, Centre for Ecological Research Institute of Biology, Eötvös University Parmenides Center for the Conceptual Foundations of Science.
Eörs Szathmáry's profile page at Centre for Ecological Research.