Category: Pop. Genetics

Summary: Mapping how adaptive-walk lengths and local-optimum structure change across K/N in NK fitness landscapes.

NK fitness landscapes model adaptation on rugged genotype spaces where each locus interacts with K other loci. This experiment asks how the structure of those landscapes changes as the interaction density K/N approaches the predicted transition near 1/e, especially in the length and variability of greedy adaptive walks.

The simulation starts from random genotypes, repeatedly takes the best single-step fitness improvement, and stops at local optima. By varying both N and K, it measures mean walk length, the number of local optima, the fitness gain over random starts, and the full shape of the walk-length distribution.

That provides a computational test of a recent theoretical transition prediction. Rather than looking only at final fitness, the experiment studies how the landscape reorganizes the whole adaptive search process.

Method: Repeated greedy adaptive-walk simulations on NK landscapes across system size N and interaction ratio K/N, followed by finite-size scaling analysis.

What is measured: Mean adaptive walk length, walk-length distribution, number of local optima, fitness gap from random starts, and transition behavior near K/N ~ 1/e.