A high-performance, concurrent Rust implementation for solving the Traveling Salesperson Problem (TSP) using a multi-caste island model genetic algorithm, 2-opt local search, and automated hyperparameter tuning.
- Island Model (Caste Architecture): Splits the population into distinct castes with varying search pressures, utilizing periodic migration to balance global exploration and local exploitation.
-
Incremental 2-Opt Local Search (Mimesis): Accelerates local optimization by evaluating move deltas in
$O(1)$ time rather than recalculating full tour costs. -
Parallel Processing: Leverages
rayonfor concurrent population evaluations, mutations, and crossovers across multi-core processors. - Automated Parameter Tuning: Includes built-in coordinate-wise hyperparameter optimization routines to fine-tune genetic operators.
- TSPLIB Support & SVG Export: Seamlessly parses standard TSPLIB-format benchmark files and automatically renders clean, styled dark-mode SVG visualizations of optimized tours.
- Language: Rust
- Concurrency:
rayon, standard library threads - Output: Custom TSPLIB parser, SVG vector rendering, structured file output