Struct Dtlz2 

pub struct Dtlz2 { /* private fields */ }

DTLZ2 (Deb, Thiele, Laumanns & Zitzler 2002): a scalable many-objective benchmark whose Pareto front is the unit hypersphere in the positive orthant (Σ fᵢ² = 1). On [0, 1]^n with M objectives and k = n − M + 1 distance variables,

g    = Σ_{i ≥ M} (xᵢ − 0.5)²
f_1  = (1+g) · cos(x₁ π/2) ··· cos(x_{M−1} π/2)
f_j  = (1+g) · cos(x₁ π/2) ··· sin(x_{M−j+1} π/2)
f_M  = (1+g) · sin(x₁ π/2)

The front is reached when g = 0 (all distance variables at 0.5); there the objective vector lies exactly on the unit sphere — the ground truth the NSGA-III test measures distance to.

Implementations

impl Dtlz2

pub fn new(m: usize, dim: usize) -> Self

m-objective DTLZ2 in dim variables (dim > m) on [0, 1]^dim. The recommended dim is m + 9 (so k = 10 distance variables).

Trait Implementations

impl MultiProblem for Dtlz2

fn dim(&self) -> usize

Number of decision variables.

fn bounds(&self) -> &[Bound]

Per-variable inclusive bounds, length MultiProblem::dim.

fn n_objectives(&self) -> usize

Number of objectives (≥ 2 for a genuine multi-objective problem).

fn objectives(&self, x: &[f64]) -> Vec<f64>

The objective vector to minimize, length MultiProblem::n_objectives.