Stabilize the API

[lars-reimann]

Stabilize the API to prepare for a 1.0.0 release. This includes checking the design, adding missing tests, and adding missing documentation. If a part of the API is not stable yet, it should be marked clearly in the documentation (#654).

Tables

• TableTransformer
• InvertibleTableTransformer
• FunctionalTableTransformer
• Discretizer
• KNearestNeighborsImputer
• LabelEncoder
• OneHotEncoder
• RangeScaler
• RobustScaler
• SimpleImputer
• StandardScaler
• SequentialTableTransformer

➡️ New release

• ColumnPlotter
• TablePlotter

➡️ New release

ML on Tables

• Dataset
• TabularDataset
• SupervisedModel
• Classifier
• AdaBoostClassifier
• BaselineClassifier
• DecisionTreeClassifier
• GradientBoostingClassifier
• KNearestNeighborsClassifier
• LogisticClassifier
• RandomForestClassifier
• SupportVectorClassifier
• Regressor
• AdaBoostRegressor
• BaselineRegressor
• DecisionTreeRegressor
• GradientBoostingRegressor
• KNearestNeighborsRegressor
• LinearRegressor
• RandomForestRegressor
• SupportVectorRegressor
• Choice (self.elements should be a property)
• ClassificationMetrics
• ClassifierMetrics (confusing to have both)
• RegressionMetrics
• RegressorMetrics (confusing to have both)

Other

• All exceptions

---

Everything that follows could be marked as experimental, and we could proceed with a 1.0.0 release.

Images

• Image
• ImageList
• ImageSize
• ImageDataset
• ConstantImageSize
• ModelImageSize
• VariableImageSize

Time Series

• TimeSeriesDataset
• ArimaModelRegressor (should be separated from the rest)

NNs

• NeuralNetworkClassifier
• NeuralNetworkRegressor (should be combined with NeuralNetworkClassifier)
• InputConversion + subclasses (should be hidden)
• AveragePooling2DLayer
• Convolutional2DLayer
• ConvolutionalTranspose2DLayer
• DropoutLayer
• FlattenLayer
• ForwardLayer
• GRULayer
• LSTMLayer
• Layer (maybe add new subclasses RecurrentLayer, ConvolutionalLayer?)
• MaxPooling2DLayer