Repository navigation
Releases: IIAS-Research/iaml
Releases · IIAS-Research/iaml
Release list
v1.1.0
1.1.0 (2026-10-07)
Features
- compile configured metric objectives (e1e3c4c)
- compile recipes into execution trees (d6a9af7)
- compose independent pipeline recipes (d43a10b)
- describe and compare recipes (c29f854)
- describe datasets without training (a4f2602)
- distinguish configured Bayesian search spaces (784f0af)
- edit recipes and component families (0443285)
- execute configured explanations on demand (4ef2684)
- expose the default pipeline as a recipe (885c7ec)
- honor explicit random seed domains (d2f03e4)
- introduce fixed values and numeric domains (73e994e)
- optimize declared training parameters (3eb4628)
- preserve metric aliases and report fold coverage (fb84207)
- reconstruct recipes as Python code (4a2ec1d)
- report resolved study snapshots (f82c744)
- restrict genetic swaps to declared alternatives (5e04552)
- track parameter policies in caches (cf7157b)
- train from attached pipeline recipes (141dda5)
Bug Fixes
- align binary curves with fitted classes (d1156f0)
- align permutation scoring with the main metric (1d94a57)
- compare final candidates on consistent evaluations (725422a)
- encode categorical features in minimal pipelines (c019e2e)
- handle open numeric bounds (97d09b4)
- honor configured MICE parameters (d6974a8)
- honor study configuration in baseline (139b24d)
- improve CatBoost hyperparameter search spaces (29f33b4)
- keep Sphinx caches outside published documentation (c81c1c3)
- make default feature selection optional (23f3583)
- preserve evaluations across sliding search stages (e4e4433)
- preserve inputs during inference (ad5062d)
- reduce survival forest memory usage (53c93f2)
- render box plots with the axes API (57d2e8b)
- respect automatic subsampling settings (1e9cc4e)
- support decision scores for binary AUROC (244304f)
- use supported MICE imputation fallback (757e296)
Performance Improvements
- skip unnecessary SVC probability calibration (321b962)
- start search with a lightweight baseline (ae75e94)
- use three folds for internal cross-validation (f534b19)
Documentation
- add basic recipe examples (227b4ca)
- add reusable strategies and study workflows (ed303a6)
- clarify stage and global search budgets (3cfde1e)
- connect guides and preserve previous URLs (c0cc94b)
- document recipe construction and editing (02fd394)
- document study analyses (7d558c5)
- introduce interactive pipeline discovery (5e38f2f)
- publish the Python example catalogue (c1ce297)
- update the README and existing user guides (bdc9521)