updated to fastmath 3
- more OLS metrices in glance
- more tidy models functions
- added function to easly retrieve datasets from Smile Github data folder
- upgraded deps
- added tidy,glance and augment for OLS
- allow classificaion of already numeric targets
- re-added some gridsearch options
- added loglikelihood calculatin for OLS
- added one-line linear regression
- adapted to tablecloth 7.x
- fixed arrity exception in
reduce-dimensions
transform - allow configuration of ppmap grain size
- fixed serialisation for svm
- added tfidf->dense
- added :tf-map-handler-fn to support pruning of the terms for tfidf
- metamorph'ed bow->tfidf reuses tf-map from :fit in :transform
- added 2 weighting schemes for tf calculation
- added 3 weighting schemes for idf calculation
- support :word-normalize-fn in count-vectorizer to configure tokenisation
- produce java 8 compatible java classes
- more docu
add 3 types of unsupervised models
- added clustering models
- added projection models
- added manifold models
- added Malli schemas to model options
- added projections
- added more model specific docu
- Upgrade to smile 2.5.0.
- Minimum workingtech.ml.dataset version is 4.00
There are fewer smile regressors and classifiers supported. XGBoost support is the
same. This requires [techascent/tech.ml.dataset "2.0-beta-56"]
or later. If you
are using dataset, xgboost, or the set of supported smile regressors and classifiers
changes to your code should be zero or minimal.
Breaking Change: This library expects tech.ml.dataset to be provided. So your project
needs both this library and [tech.ml.dataset "2.0-beta-49"]
. This is to reduce the
number of spurious releases.