Skip to main content

Crate u_metaheur

Crate u_metaheur 

Source
Expand description

Domain-agnostic metaheuristic optimization framework.

Provides generic implementations of common metaheuristic algorithms:

  • Genetic Algorithm (GA): Population-based evolutionary optimization with pluggable selection, crossover, and mutation operators.
  • BRKGA: Biased Random-Key Genetic Algorithm — the user implements only a decoder; all evolutionary mechanics are handled generically.
  • Simulated Annealing (SA): Single-solution trajectory optimization with pluggable cooling schedules.
  • ALNS: Adaptive Large Neighborhood Search — destroy/repair operators with adaptive weight selection.
  • Dispatching: Generic priority rule composition engine for multi-rule item ranking.
  • CP (Constraint Programming): Domain-agnostic modeling layer for constrained optimization with interval, integer, and boolean variables.
  • Tabu Search (TS): Single-solution trajectory optimization using short-term memory (tabu list) to escape local optima.
  • Variable Neighborhood Search (VNS): Systematic neighborhood switching for escaping local optima via diversified perturbation.

§Architecture

This crate sits at Layer 2 (Algorithms) in the U-Engine ecosystem, depending only on u-numflow (Layer 1: Foundation). It contains no domain-specific concepts — scheduling, nesting, routing, etc. are all defined by consumers at higher layers.

Modules§

alns
Adaptive Large Neighborhood Search (ALNS) framework.
brkga
Biased Random-Key Genetic Algorithm (BRKGA).
cp
Constraint Programming (CP) framework.
dispatching
Generic priority rule composition framework.
ga
Genetic Algorithm framework.
sa
Simulated Annealing (SA).
tabu
Tabu Search (TS).
vns
Variable Neighborhood Search (VNS).