HeuristicFlow v1.6.3

Updates

1. Add interface cmake targets
2. Install like a cmake package
3. Configure Eigen3 automatically
4. Fixed an error in SimpleMatrix

HeuristicFlow v1.6.2.1

This version has no new algorithms but some slight new tools.

Updates

• You can convert a double to its bianry form with function constexpr heu::binCode heu::encode(double), where heu::binCode is a 64 btis unsigned enumeration type. For inverse conversion, use constexpr double heu::decode(heu::binCode). These two functions computes at compile time.
• New box constraints : heu::FixedDiscreteBox and heu::FixedContinousBox. You should pass the range via template parameter so that the range is fixed at compile time. For C++17, floating-point can't be passed through template parameters, so we have heu::FixedContinousBox17, which receives range data in heu::binCodes.

HeuristicFlow v1.6.2

Updates:

• Add 10 selection methods for SOGA : RouletteWheel, Tournament, MonteCarlo, Probability, LinearRank, ExponentialRank, Boltzmann and EliteReserved. Selectiong method can also be determined at runtime.
• Matrix types in SimpleMatrix module can evulate between different types.
• Box constraint in EAGloable module is rewritten.
• Add infinite box constraint (initialize by gaussian distribution).
• The PSO module has its own special box constraint.

Full Changelog: v1.6.1...v1.6.2

HeuristicFlow v1.6.0

Updates :

• New module : Atomic Orbit Search (AOS) and a new algorithm of the same name
• 19 template test functions for single-objective solvers (4 scalable and 15 nonscalable) in module EAGlobal
  • You can call them like heu::testFunctions<Eigen::Array2d,double>::eggHolder.
• heu::array_traits<class T> : New metafunction that are able to tell attributes of a general vector class like type, compile-time-size, element-type and whether a Eigen class or not.
• Box constraints have new member fuction applyConstraint(...) now.

Changes :

• Depends on C++17 now.

HeuristicFlow v1.5.0

Updates:

• Add PSODefaults.

Changes:

• Generate document by doxygen
• Optimized NSGA2's performance.
• Add denpendent : libEigen
• ...

HeuristicFlow v1.4.1

Updates:

• Compability to C++14
• Versatile box constraint for all kinds of real, binary and symbolic encoding. You can assign much more variables at compile time.
• More default GA operators under box constriant : iFunNd,iFunXd, iFunNb, iFunXb, iFunNs, iFunXs, mFun_d, mFun_b and mFun_s.
• PSO module supports compile time initializpr now.

Changes:

• Macro in source file Heu_MAKE_GLOBAL no longer required.

HeuristicFlow v1.4.0

Updates:

• New module : EAGlobal
• Box constraint for GA
• min and max function with multiple parameters

Changes:

• GA now supports compile-time function binding
• NSGA2 uses an adaptively normalized congestion now
• Renamed OtGlobal after HfGlobal

HeuristicFlow v1.3.3

This project has been renamed as HeuristicFlow, while previous namespace name OptimT was renamed as Heu.

Updates:

• New MOEA: NSGA3
• New generic function to compute matrix product and inverse matrix using Heu::MatrixDynamicSize and Heu::MatrixFixedSize.
• MatrixMap.
• 2 methods to convey floating point numbers in template under C++20. (Heu::DivCode and Heu::PowCode)

Changes:

• Macro OptimT_DO_PARALLELIZE is replaced by Heu_USE_THREADS.
• Macro OptimT_NSGA2_RunTimeObjNum_MaxRTObjNum is replaced by Heu_MOGA_MaxRunTimeObjNum.
• Macro OptimT_USE_EIGEN,OptimT_GA_USE_EIGEN and OptimT_PSO_USE_EIGEN are no longer required. Partial specializations for Eigen will be activated once you included Eigen before this lib.

Bugs Fixed:

• Fixed the bug that the size of MOGA's Pareto frontier exceeds the population size.

OptimTemplates v1.3.2

Updates:

• Fresher API : replace std::tuple<Args...> with single custom type.
• Replace otherOptFun with virtual function virtual void customOptAfterEachGeneration().
• Testing version of NSGA3.

Bugs fixed:

• Fixed the bug that MatrixDynamicSize failed to compile in debug mode -- static member are now encapsulated in static function.

Changes:

• When no other parameters provided, Args_t won't appear in initialize and fitness functions

OptimTemplates v1.3.1

Updates:

• NSGA2 now supports using Eigen's array types (Eigen::Array<double,...>) as fitness value with vectorized implementation of selection operation.
• NSGA2 now supports working with runtime objective numbers.

Changes:

• New classes: MOGAAbstract, MOGABase, NSGABase, NSGA2Base (All of them are used to reduce code duplication while supporting runtime objective numbers and boosting with eigen arrays).