The (non-Gaussian) structure of fitness distributions
speaker: Matteo Smerlak (Max Planck Institute for Mathematics in the Sciences)abstract: At about the same time he was studying the basic equations of evolutionary dynamics, Ronald Fisher co-founded extreme value theory, the statistical analysis of extreme events. Fisher appears to never have made an explicit link between these two contributions, yet their mathematical structure is very similar: in extreme value theory, you pick the largest member of a large sample; in evolution, natural selection preferably picks the fittest member of a large population. In this talk I will show that, like distributions of sample maxima, fitness distributions are subject to statistical universality. I will derive the corresponding limiting distributions and argue that these distributions (flipped gamma distributions) provide a better description of evolutionary dynamics in fitness space than Gaussian traveling waves.
timetable:
Wed 3 Apr, 14:00 - 15:00, Aula Dini
<< Go back