Overlaps in spin glasses and models of evolution

A measure of similarity or overlap between different configurations or individuals appears in many complex systems such as spin glasses, evolving populations, optimization problems and neural networks. In the context of spin glasses, the concept was introduced by Edwards and Anderson 50 years ago and gives a measure of similarity between pairs of spin configurations. In the context of evolving populations, the overlap measures the similarity between genomes of individuals. In general the overlaps fluctuate and have non-trivial statistics. Giorgio Parisi's replica theory of mean field spin glasses predicted a universal form for the overlap statistics. In this talk I will compare these predictions with the statistics for different models of disordered systems and for evolving populations.

Please note if you join through the Microsoft Teams link, your name and profile photo appear on the recording. If you have any issues with this, please contact hctp@ed.ac.uk.