Module replacement 

Survivor-selection / replacement operators.

Each function takes the current generation’s population and fitness plus the offspring population and fitness, and returns the (population, fitness) pair for the next generation.

The helpers work on host-side fitness slices for selection logic and lift winners back to the device via Tensor::select.

Functions

elitist

Elitist replacement: keep the k best members of the current generation and fill the rest from offspring, preserving the best offspring first.

generational

Replace the entire current generation with offspring (no elitism).

mu_comma_lambda

(μ, λ) replacement: discard parents; keep the μ best of the λ offspring. Requires lambda >= mu.

mu_plus_lambda

(μ + λ) replacement: merge the μ parents with the λ offspring and keep the μ lowest-fitness members overall.