Overview
HeuristicLib is a .NET library for building, running and studying heuristic optimization algorithms. It provides ready to use algorithms for common searches and typed building blocks for problems that need a custom approach.
Use it when an answer must be found among many possible candidates and checking a candidate is easier than deriving the best answer directly. Typical examples include scheduling, routing, parameter tuning and symbolic regression.
The model in one sentence
An algorithm uses operators to explore the search space of a problem and yields search states that show its progress.
Problem
Search space
Evaluation
Objective
Algorithm configuration
Operators
Population and budgets
Random source
Recorded root seed
Algorithm run
Fresh execution state
Search states
Evaluated candidates and progress counters
The parts stay separate. Compare algorithms on the same problem, swap a mutation operator or repeat a configuration with different random seeds.
What is included
- Population based algorithms including genetic algorithms, evolution strategies and NSGA-II
- Local search algorithms including hill climbing
- Search spaces for real, integer, Boolean and permutation vectors
- Reusable creation, crossover, mutation and selection operators
- Single objective and multiobjective evaluation
- Repeatable experiments, parameter grids and analysis helpers
- Specialized support for machine learning, symbolic regression and symbolic expressions
- Source based Python integration through pythonnet
HeuristicLib targets .NET 10. The package is currently prerelease software, so expect APIs to evolve between versions.
A typical workflow
- Describe valid candidates with a search space.
- Define how the problem evaluates a candidate.
- Choose an algorithm and compatible operators.
- Run with an explicit random seed.
- Inspect the final result or stream intermediate states.
- Repeat runs to compare configurations rather than trusting one stochastic result.
The getting started guide builds a complete run. Read core concepts for the parts used by that example.