Rademacher Complexity and Models of Group Competition
12 days ago
- #evolutionary biology
- #Rademacher Complexity
- #group selection
- Rademacher Complexity and mean field approximations are applied to kin and group selection models in evolutionary biology.
- Group selection theory is presented as a mean field approximation, simplifying complex interactions by averaging effects.
- Kin selection theory, formalized by Hamilton's Rule (rB > C), explains altruism through genetic relatedness and cost-benefit analysis.
- Nowak, Tarnita, and Wilson (NTW) argue that kin selection is limited and propose group selection as an alternative model.
- Mean field theories in physics, like those by Curie and Weiss, inspired NTW's approach by averaging interactions in complex systems.
- NTW's model treats worker ants as 'robots' of the queen, ignoring individual interactions for colony-level dynamics.
- Rademacher Complexity measures model richness, comparing how well kin and group selection models generalize beyond training data.
- Group selection models show lower Rademacher Complexity for group dynamics, while kin selection models excel in individual interactions.
- Competition between groups fosters cooperation within groups, aligning individual incentives with collective success.
- The essay concludes that kin and group selection are complementary, each useful at different scales of biological and social organization.