History¶

v0.0.11-alpha¶

This is the initial release of iwopy - ready for testing.

So far not many models have been transferred from the Fraunhofer IWES in-house predecessor iwes-opt, they will be added in the following versions. We are just getting started here!

Enjoy - we are awaiting comments and issues, thanks for testing.

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.0.11

v0.0.12-alpha¶

• Benchmarks:

  • New benchmark rosenbrock added

• Tests:

  • Extended tests for pygmo and pymoo

• Fixes:

  • Bugs fixed with discretization

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.0.12

v0.0.13-alpha¶

• Wrappers:

  • Introducing SimpleObjective and SimpleConstraint, for functions of simple numeric types like f(x, y, ...)

• Examples:

  • New example: electrostatics, solvable with IPOPT, GA, PSO

• Notebooks:

  • Work on examples.ipynb: Adding section on simple function minimization

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.0.13

v0.0.14-alpha¶

• Notebooks:

  • Work on examples.ipynb: Adding electrostatics example

• Core:

  • OptResults can now be printed

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.0.14

v0.0.15-alpha¶

• Utils:

  • RegularDiscretizationGrid now with interpolation parameter: None, nearest or linear

• Wrappers:

  • New wrapper: LocalFD, Local Finite Difference. Calculates function derivatives by applying local finite difference rules. Currently orders 1 and 2 are implemented, more could be added later.

• Optimizers:

  • New optimizer: GG, Greedy Gradient. A pretty straight forward gradient based local optimizer that projects out (or reverses) directions of constraint violation. Mostly implemented for the purpose of testing gradients, but maybe worth a try also for problem solving.

• Examples:

  • electrostatics: New optional constraint MinDist, forcing charges to keep a minimal distance. Also adding a script that demonstrates how to solve this problem using NLOPT via pygmo, and another for solving the problem using the GG algorithm.

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.0.15

v0.1.0-alpha¶

• Bug fixes

• Updated documentation

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.0

v0.1.1-alpha¶

• Tests:

  • pygmo: Increased ngen and npop, intended for osx systems to pass tests

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.

v0.1.2-alpha¶

• Tests:

  • pygmo: Fixing test such that it passes on all systems

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.2

v0.1.3-alpha¶

• Core:

  • Support for multi objective optimizations

  • Splitting OptResults into SingleObjOptResults and MultiObjOptResults

• Interfaces:

  • pymoo: Adding nsga2

• Examples:

  • New: multi_obj_chain, demonstrating multi objective optimization using pymoo’s NSGA2

• Notebooks:

  • New: multi_obj_chain.ipynb, demonstrating multi objective optimization using pymoo’s NSGA2

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.3

v0.1.4-alpha¶

• Interfaces:

  • Improved: pymoo now supports mixed int/float problems and algorithm MixedVariableGA

  • Changed: pymooalgorithms now selectable via same name as in package, i.e., GA, PSO, NSGA2, MixedVariableGA

• Examples:

  • New: mixed, demonstrating how to solve a mixed int/float problem with pymoo

• Notebook:

  • New: mixed, demonstrating how to solve a mixed int/float problem with pymoo

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.4

v0.1.5-alpha¶

• Small bug fixes

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.5

v0.1.6-alpha¶

• Interfaces

  • New interface to scipy: Calling the scipy.optimize.minimize function

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.6

v0.1.7-beta¶

• Utils

  • New import_module utility, dynamically loading packages and modules

• Interfaces

  • Improved dynamic imports of pygmo and pymoo

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.7

v0.1.8-beta¶

• Bug fixed in setup.cfg

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.8

v0.1.9-beta¶

• Bug fixes

  • Fixes for bugs in tests for Python 3.7 and 3.11

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.1.9

v0.2-beta¶

• Interfaces:

  • Improved pymoo: Dynamical derivation from classes from optional package import

• Documentation:

  • Completely new sphinx style

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.2

v0.2.1¶

• Documentation improvements

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.2.1

v0.2.2¶

• Restricting numpy to <2, due to dependencies

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.2.2

v0.2.3¶

• Recovering support for Python 3.7

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.2.3

v0.3¶

• removing restriction numpy <2, since now latest numpy is supported again by indirect dependency autograd

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.3

v0.3.1¶

• Added factories for allowing automatized setups

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.3.1

v0.4.0¶

• Dropping support for Python 3.8

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.4.0

v0.5.0¶

• Core:

  • Added optimization pipelines through the new abstract Pipeline base class.

  • Added native optimizer callbacks with normalized intermediate states for GG, SLSQP, SciPy, pymoo, and PyGMO. Callback validation and finalization are now consistent, callback-free runs avoid adapter overhead, and unsupported PyGMO IPOPT callbacks fail explicitly.

  • Unified problem summaries across optimizer output.

• Optimizers:

  • Added SLSQP for constrained single-objective problems, with automatic dimensionless scaling, vectorized iwopy Jacobians, evaluation-free progress output, and reuse of cached function values in LocalFD.

  • Hardened GG for zero and non-finite gradients, general constraint bounds, stalled constraints, hard iteration limits, and convergence edge cases.

• Interfaces:

  • Extended pymoo factories with DE, NSGA3, and CMAES, and improved integer offspring handling, reusable configurations, and bounded single-objective termination.

  • Expanded PyGMO algorithm support and made IPOPT option handling more robust.

• Problems:

  • Added DiscreteLocalMoveOptProblem for applying one small discrete move per wind turbine.

• Examples:

  • Added runnable SLSQP and pymoo examples for general constraints and callbacks, including live progress and objective-history plots.

• Bugs:

  • Fixed lower-bound second-order finite differences in LocalFD, maximization gradient orientation in PyGMO, and callback leakage after PyGMO solves.

  • Fixed regular-grid index and interpolation error handling, SciPy mixed-variable bounds and cache eviction, disabled pymoo problem-result storage access, and empty PyGMO gradient sparsity handling.

  • Rejected boolean values for integer-only GG iteration settings and GG or SLSQP memory sizes.

  • Fixed package-version fallback lookup on Python versions without tomllib.

• Dependencies:

  • Added the iwopy[scipy], iwopy[pymoo], iwopy[pygmo], and iwopy[opt] extras; dropped Python 3.9 and raised runtime, PyGMO, test, and documentation dependency floors.

• Development:

  • Added package-wide type annotations and strict mypy checking, enforced by pre-commit.

  • Standardized API docstrings and Sphinx configuration on the foxes NumPyDoc style, with a Python 3.10 TOML fallback for documentation builds.

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v0.5.0

v1.0.0¶

• Python versions:

  • Added Python 3.14 support; Python 3.10 remains the minimum supported version.

• Core:

  • Jacobian assembly and optimizer caches now use SciPy CSR matrices instead of dense arrays. Use Problem.get_jacobian() and calc_jacobian() for matrix derivatives; scalar gradients and backend-required native names retain their meaning.

  • Added iwopy.create_optimizer to select a backend family and algorithm, including a PyGMO-backed nlopt family. The root and core exports are the same function; public optimizer class names are unchanged.

  • Pipelines support restart result seeding for selected stage ranges and clear running state before finalization after stage-reported failures.

• Wrappers:

  • LocalFD supports bounded vectorized finite-difference batches through max_population_size and avoids evaluating analytically differentiated components during numerical Jacobian completion.

• Optimizers:

  • GG caps fresh finite-difference steps without extra backtracking evaluations, aligns progress output with completed-iteration callbacks, and keeps a feasible iterate when every trial violates constraints.

  • SLSQP retains its best fully feasible point independently of convergence, supports inward margins and bounded restoration with explicit positive restoration_scales, and starts objective-progress patience only after feasibility. Restoration merit and derivatives use consistent normalization; bound-certified row reduction preserves full feasibility checks.

• Interfaces:

  • Gradient-capable SciPy minimizers use cached, population-vectorized Jacobians and grouped general constraints for continuous variables.

  • Added SciPy differential evolution with mixed integer/float variables, cached nonlinear constraint evaluations, callbacks, and per-dimension population sizing.

  • Added lazy optional Nevergrad, Optuna, and DEAP interfaces for bounded single-objective problems, general constraints, and evaluated-candidate results. The evolutionary adapter exposes generation limits.

  • Added the lazy optional cma interface for pycma CMA-ES with native augmented-Lagrangian constraints, callbacks, finalized results, and termination diagnostics; it supports scalar and population evaluation.

  • pymoo run limits compose with custom termination criteria

  • pygmo reads sparse fitness-Jacobian entries without densification.

• Benchmarks:

  • Published QuadraticProblem, ParetoProblem, and MixedProblem through iwopy.benchmarks, replacing the optimizer examples’ local shared-case module.

• Examples and documentation:

  • Added an optimizer catalog with 81 runnable, main-guarded factory-call scripts using installed-package imports without path setup. Entries are grouped by objective, constraint, and local/global capabilities, with practical recommendations and links to class APIs.

  • Added constrained NSGA2/NSGA3 Pareto examples with explicit weighted selection. All maintained examples and notebooks use the root optimizer factory and current constructor and run-limit controls; maintained FOXES Optimization callers were migrated too.

  • Reworked optimizer setups with individually annotated controls, short demo budgets, and production starting suggestions. Local PyGMO solvers use one starting point; native iwopy optimizers are identified as such.

  • Expanded the Vectorized Interfaces overview to cover all adapters and distinguish search, Jacobian, and batch-fitness vectorization. Added shared parameter and termination support tables.

  • Corrected chain constraint counts, callback failure semantics, and the gradient optimizer architecture link; aligned token-review guidance with the policy keeping documentation checks outside pytest.

  • Consolidated these release notes by area, separating user-facing changes and breaking APIs from contributor workflow details.

Full Changelog: https://github.com/FraunhoferIWES/iwopy/commits/v1.0.0