v2023.3.10

[2023.3.9]

Changed

• Update dependencies

Fix

• ValueError: X and y both have indexes, but they do not match.

1.3.10

[1.3.10]

Fix

• TypeError in data_prepare Outliers filter

[1.3.9]

ADD

• Up score AutoML (Blend best top5 models in AutoML)

1.3.8

ADD

• optimization DataPreproc parametrs in BestSingleModel
• rebuild AutoML pepline (light version)

Fix

• target encodet only cat features

1.3.7

Fix

• target encoder in model.opt

1.3.6

[1.3.6]

ADD

• add dosc on CV

[1.3.5]

Fix

• Fix nans in targetencoder in CV

[1.3.4]

ADD

• Target Encoding in CrossValidation
• DenoisingAutoencoder in DataPrepare
• Docs

1.3.1

[1.3.1]

Fix

• Fix import - add loguru and psutil in requirements.txt

[1.2.28]

ADD

• Advanced Logging (logs in .automl-alex_tmp/log.log)
• Class Optimizer
• Pruner in optimizer
• connection with optuna-dashboard (run > optuna-dashboard sqlite:///db.sqlite3 )
• NumericInteractionFeatures Class in data_prepare

1.1.25

[1.2.25]

Fix

• Fix save & load in AutoML

ADD

• Metod .score() and .fit_score() in Models
• Class CrossValidation() examples in ./examples/03_Models.ipynb

1.2.23

Big Update:

A big update that changes the logic of work

NEW

• Now processing the dataset is separated from the model for ease of use when you want to process the dataset yourself
• Separate transform allows us to save and transfer processing to new data (work in production)

ADD

• Save & Load processing
• Save & Load model
• Reduce memory usage processing
• Detect and remove outliers

AutoML itself now rebuild, so it is possible to drop the score. If the metric is important to you, you can use the previous version.

Work on progress...

1.01.11

Fix

• score_cv_folds fix in ModelsReview
• normalization

0.11.24

ADD

• multivariate TPE sampler. This algorithm captures dependencies among hyperparameters better than the previous algorithm

Fix

• "ValueError non-broadcastable output operand..." in AutoMLRegressor