Post-catastrophe evolution follows Levenberg-Marquardt optimization, not gradient ascent
A new theory of natural selection following catastrophic environmental change shows that mean fitness recovery follows an inverse-time law, with a prefactor set by the number of traits under selection. Testing against experimental E. coli fitness landscapes after antibiotic perturbation reveals that post-catastrophe adaptation does not follow gradient ascent on a fitness landscape but instead matches the Levenberg-Marquardt optimization algorithm, suggesting evolution becomes 'optimistic' rather than greedy near fitness peaks.