Crate moors 

moors

Multi‑Objective Optimization in Pure Rust
Fast, extensible evolutionary algorithms with first‑class ndarray support.

---

Overview

moors provides a battery of evolutionary algorithms for solving multi‑objective optimization problems. The core goals are:

• Performance – minimal allocations, Rayon‑powered parallel loops where it matters.
• Extensibility – every operator (sampling, crossover, mutation, selection, survival) is pluggable via pure Rust traits.

Currently implemented algorithms

Family	Algorithms
NSGA	NSGA‑II, NSGA‑III, RNSGA‑II
SPEA	SPEA‑2
Others	AGE‑MOEA, REVEA (WIP)

Quick start

use ndarray::{Array1, Array2, Axis, stack};

use moors::{
    algorithms::{AlgorithmError, Nsga2Builder},
    duplicates::ExactDuplicatesCleaner,
    operators::{SinglePointBinaryCrossover, BitFlipMutation, RandomSamplingBinary},
};

// ----- problem data (0/1 knapsack) -------------------------------------
const WEIGHTS: [f64; 5] = [12.0, 2.0, 1.0, 4.0, 10.0];
const VALUES:  [f64; 5] = [ 4.0, 2.0, 1.0, 5.0,  3.0];
const CAPACITY: f64 = 15.0;

/// Multi‑objective fitness ⇒ [−total_value, total_weight]
fn fitness(pop_genes: &Array2<f64>) -> Array2<f64> {
    let w = Array1::from_vec(WEIGHTS.into());
    let v = Array1::from_vec(VALUES.into());
    let total_v = pop_genes.dot(&v);
    let total_w = pop_genes.dot(&w);
    stack(Axis(1), &[(-&total_v).view(), total_w.view()]).unwrap()
}

/// Single inequality constraint ⇒ total_weight − CAPACITY ≤ 0
fn constraints(pop_genes: &Array2<f64>) -> Array1<f64> {
    let w = Array1::from_vec(WEIGHTS.into());
    pop_genes.dot(&w) - CAPACITY
}

fn main() -> Result<(), AlgorithmError> {
    let mut algo = Nsga2Builder::default()
        .fitness_fn(fitness)
        .constraints_fn(constraints)
        .sampler(RandomSamplingBinary::new())
        .crossover(SinglePointBinaryCrossover::new())
        .mutation(BitFlipMutation::new(0.5))
        .duplicates_cleaner(ExactDuplicatesCleaner::new())
        .num_vars(5)
        .population_size(100)
        .crossover_rate(0.9)
        .mutation_rate(0.1)
        .num_offsprings(32)
        .num_iterations(200)
        .build()?;

    algo.run()?;
    Ok(())
}

Module layout

• algorithms – high‑level algorithm builders
• operators – sampling, crossover, mutation, selection, survival
• genetic – core data types (Individual, Population, …)
• evaluator – fitness + constraints evaluation pipeline
• random – pluggable RNG abstraction
• duplicates – duplicate‑handling strategies

---

Re-exports

pub use algorithms::AgeMoea;

pub use algorithms::AgeMoeaBuilder;

pub use algorithms::AlgorithmBuilder;

pub use algorithms::AlgorithmBuilderError;

pub use algorithms::AlgorithmError;

pub use algorithms::GeneticAlgorithm;

pub use algorithms::Ibea;

pub use algorithms::IbeaBuilder;

pub use algorithms::InitializationError;

pub use algorithms::Nsga2;

pub use algorithms::Nsga2Builder;

pub use algorithms::Nsga3;

pub use algorithms::Nsga3Builder;

pub use algorithms::Revea;

pub use algorithms::ReveaBuilder;

pub use algorithms::Rnsga2;

pub use algorithms::Rnsga2Builder;

pub use algorithms::Spea2;

pub use algorithms::Spea2Builder;

pub use duplicates::CloseDuplicatesCleaner;

pub use duplicates::ExactDuplicatesCleaner;

pub use duplicates::NoDuplicatesCleaner;

pub use duplicates::PopulationCleaner;

pub use evaluator::ConstraintsFn;

pub use evaluator::EvaluatorError;

pub use evaluator::FitnessFn;

pub use evaluator::NoConstraints;

pub use genetic::Individual;

pub use genetic::IndividualMOO;

pub use genetic::IndividualSOO;

pub use genetic::Population;

pub use genetic::PopulationMOO;

pub use genetic::PopulationSOO;

pub use operators::selection;

pub use operators::survival;

pub use operators::AgeMoeaSurvival;

pub use operators::ArithmeticCrossover;

pub use operators::BitFlipMutation;

pub use operators::CrossoverOperator;

pub use operators::DanAndDenisReferencePoints;

pub use operators::DisplacementMutation;

pub use operators::ExponentialCrossover;

pub use operators::FrontsAndRankingBasedSurvival;

pub use operators::GaussianMutation;

pub use operators::InversionMutation;

pub use operators::MutationOperator;

pub use operators::Nsga2RankCrowdingSurvival;

pub use operators::Nsga3ReferencePointsSurvival;

pub use operators::OrderCrossover;

pub use operators::PermutationSampling;

pub use operators::RandomSamplingBinary;

pub use operators::RandomSamplingFloat;

pub use operators::RandomSamplingInt;

pub use operators::RandomSelectionMOO;

pub use operators::RankAndScoringSelectionMOO;

pub use operators::ReveaReferencePointsSurvival;

pub use operators::Rnsga2ReferencePointsSurvival;

pub use operators::SamplingOperator;

pub use operators::ScrambleMutation;

pub use operators::SelectionOperator;

pub use operators::SimulatedBinaryCrossover;

pub use operators::SinglePointBinaryCrossover;

pub use operators::Spea2KnnSurvival;

pub use operators::StructuredReferencePoints;

pub use operators::SurvivalOperator;

pub use operators::SwapMutation;

pub use operators::TwoPointBinaryCrossover;

pub use operators::UniformBinaryCrossover;

pub use operators::UniformBinaryMutation;

pub use operators::UniformRealMutation;

pub use operators::evolve::EvolveError;

pub use random::MOORandomGenerator;

pub use random::NoopRandomGenerator;

pub use random::RandomGenerator;

pub use random::TestDummyRng;

Modules

algorithms

duplicates

duplicates – Keeping the Gene Pool Diverse

evaluator

evaluator – From Genes to Population

genetic

genetic – Core Data Structures

non_dominated_sorting

operators

operators – Building Blocks for Evolution

random

random – Unified RNG Abstraction

Macros

__constraints_helper

Concatenate one or more already-wrapped constraint closures (inequalities or equalities) into a single evaluator closure returning an Array2.

__eq_helper

Wrap a single constraint function as an equality (|g(genes)| - ε ≤ 0, ε = 1e-6).

define_algorithm_and_builder

fitness_fn

Build a composite fitness-evaluation closure from one or more scalar / vector fitness functions.

impl_constraints_fn

Defines a struct and implements [moors::ConstraintsFn] for it.

Functions

cross_euclidean_distances

Computes the cross squared Euclidean distance matrix between data and reference using matrix algebra.