Enum DeVariant 

pub enum DeVariant {
    Rand1Bin,
    Best1Bin,
    CurrentToBest1Bin,
    Rand2Bin,
    Rand1Exp,
}

Mutation + crossover variant for differential evolution.

Convergence caveats

Not every variant converges to machine precision on every landscape within the same budget. On unimodal landscapes like Sphere, Best1Bin and CurrentToBest1Bin tend to converge prematurely: the population collapses around the current best before the differential search has fully explored, and the per-generation variance F · (x_{r2} − x_{r3}) shrinks to zero. Classical DE literature documents this as the core trade-off of best-biased variants. The crate’s integration tests therefore only require strong reduction from the random baseline for those variants, not optimality — see algorithms::de::tests::all_variants_converge_on_sphere_d10 for the per-variant tolerance choice.

Variants

Rand1Bin

x_{r1} + F · (x_{r2} − x_{r3}), binomial crossover. Balanced exploration / exploitation; reaches machine precision on Sphere within a few hundred generations.

Best1Bin

x_{best} + F · (x_{r2} − x_{r3}), binomial crossover.

Strong exploitation — the mutation base is always the current best, so the population concentrates quickly. Prone to premature convergence on landscapes where the current best is far from the global optimum; on Sphere-D10 with 500 gens this variant stalls around best_fitness ≈ 1 while Rand1Bin reaches < 1e-20.

CurrentToBest1Bin

x_i + F · (x_{best} − x_i) + F · (x_{r1} − x_{r2}), binomial.

Hybrid of the current individual and the best-so-far. Still prone to premature convergence because the F · (x_{best} − x_i) term dominates once the population is near the best. Useful on multimodal landscapes where pure-best variants get stuck in local basins, less useful on Sphere.

Rand2Bin

x_{r1} + F · (x_{r2} − x_{r3}) + F · (x_{r4} − x_{r5}), binomial. Higher variance than Rand1Bin thanks to two difference vectors; converges on Sphere but more slowly.

Rand1Exp

x_{r1} + F · (x_{r2} − x_{r3}), exponential crossover. Identical mutation to Rand1Bin, different crossover mask shape. Performance comparable to Rand1Bin in practice.

Trait Implementations

impl Clone for DeVariant

fn clone(&self) -> DeVariant

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for DeVariant

impl Debug for DeVariant

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for DeVariant

impl PartialEq for DeVariant

fn eq(&self, other: &DeVariant) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for DeVariant