Regensburg 2022 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 7: Statistical Physics of Biological Systems 1 (joint session BP/DY)
DY 7.6: Vortrag
Montag, 5. September 2022, 16:45–17:00, H16
Playing it safe: information constrains collective betting strategies — •Philipp Fleig1,2 and Vijay Balasubramanian2 — 1Max Planck Institute for Medical Research, 69120 Heidelberg, Germany — 2Department of Physics & Astronomy, University of Pennsylvania, Philadelphia, PA 19104, USA
Risk is an inherent part of life and biological functions are partly shaped by the need to reduce risk. Broadly, risk arises from stochastic interactions of an organism with its environment. Every time an organism displays a particular response or behaviour (e.g. expresses a phenotype or exhibits a certain immune response), it is placing a bet with potential impact on its biological fitness. The more precisely the statistics of the environment are known to the organism, the more successfully bets can be placed. However, an organism typically has limited information about the statistics of the environment. This limitation should be accounted for in the adaptation of biological functions to the environment. We develop a theoretical principle where information geometric model complexity guides stochastic biological functions towards less risky betting strategies. In the framework of Bayesian inference, we show that given finite information about the environment, there is an optimally safe adaptation strategy set by the Bayesian prior. Furthermore, in a toy model of stochastic phenotypic switching by bacteria, we demonstrate how the implementation of our principle of ``playing it safe" increases the fitness (population growth rate) of the bacterial collective. We suggest that the principle applies broadly to problems of adaptation, learning and evolution.