Removed generator-based implementation and switched to newer pipeline implementation for data generation tasks. ML still has to be switched.

Changelog: feature

Changelog: added

Introduced model-based data pipeline algorithm and switched to more configurable data validation. More to come.

Changelog: feature

Export the serialised models of predictive and density to allow other plugins to use the information.

* Languages: 2025.4.0-SNAPSHOT
* Maven: 2025.4.0.rc1.SNAPSHOT

* Refactored parts of the statistics API and made it part of `de.evoal.optimisation.api.model`. The result was renamed to `Iteration`.
* Renamed component `individuals-per-generation` to `candidates-per-iteration`
* Renamed component `best-individual-per-generation` to `best-candidate-per-iteration`
* Adapted all examples to reflect the renaming
* Implemented a `StoppingCriterion` concept for optimisation algorithms.
* Added an initial `iteration-count` component that allows terminates the optimisation algorithm after a specified number of iterations.
* Adapted EA to use the stopping criterion.
* Adapted EA examples to use stopping criterion correctly. 
 
Changelog: feature

Changelog: added

Lauren Paul authored 3 months ago and Bernhard Johannes Berger committed 3 months ago

Changelog: bug

Changelog: example

Closes #137

Changelog: other

Lauren Paul authored 4 months ago and Bernhard Johannes Berger committed 4 months ago

* Hill Climbing
* Simulated Annealing

Changelog: added
Issue: #131 #134

Lauren Paul authored 4 months ago and Bernhard Johannes Berger committed 4 months ago

There was an issue with the `Optimisation Distance Component` due to an attribute naming conflict. In this fix, we renamed the attribute `name` to `var` as `name` is a reserved name in our DSLs.

Changelog: bug
Issue: #135

We changed all version numbers from 2024.10.0 to 2025.4.0, as we are now working on the release candidate for the next official release.

Support integer literals in range expression of for loops.

## `generator`
Support integer literals in range expressions of for loops and allow variable usage in count part of write statement.  

The changes are downwards compatible and resolve #120 "Dynamic sample size in for-loops".

Changelog: changed

Implements a new pipeline component that uses a surrogate model to predict data.

Solves #117

Adding a script for automatically deleting unused packages from the package registry. This script is triggered by a scheduled pipeline. 


Resolves !34 "Automatically remove packages for deleted branches"

This commit resolves issue #116 "ClassCastException when setting integer value for σ in noise-data component."

Reworked the handling of DSL files for developers. The `LanguageHelper` component was renamed to `AttributeHelper` and is more powerful than the old version. The function `lookup` automatically does default value handling (using default values of the attribute definition if the attribute is missing in the DSL) and automatic type casts. Thus, `int` values are now automatically casted into `double` values if the attribute definition states that the attribute is of type `real`.

Changelog: changed

Lauren Paul authored 7 months ago and Bernhard Johannes Berger committed 7 months ago

1) `Requirements.requireProbability` checked the wrong upper bound
2) `constraint`, `correlated` and `ranged` statistics used wrong CDI names
3) The `correlated` statistics writer also logged the fitness values of other fitness calculation strategies. This was removed.

Closes #93

Changelog: bug

Closes #86

Added regression tests for all surrogate models provided by `surrogate.simple` and `surrogate.smile`. Fixed minor bugs that prevented regression tests to work.

Closes #84
Resolve "Create regression tests for surrogate functions"