Welcome to bosk’s documentation!#
Bosk is a framework for Deep Forest construction.
Following common principle of deep neural network frameworks, we consider models as general computational graphs with some additional functionality, in contrast to defining strictly layerwise structure. In bosk a Deep Forest structure corresponds to two separate computational graphs: one for fitting (training) and one for transforming (predicting).
This framework helps to construct new Deep Forests avoiding writing error prone routine code, and provides tools for pipeline execution and debugging, as well as a wide set of ready to use building blocks. It supports both fully manual pipeline and automatical layerwise Deep Forest building.
Quick example#
For example, to define Deep Forest with one layer, consisting of two forests (Random Forest and Extremely Randomized Trees), which output probabilities are concatenated with input feature vector and passed to the final forest, the following code could be used:
# make a pipeline
b = FunctionalPipelineBuilder()
# placeholders for input features `x` and target variable `y`
x = b.Input('x')()
y = b.TargetInput('y')()
# random forests
random_forest = b.RFC(max_depth=5)
extra_trees = b.ETC(n_estimators=200)
# concatenation
cat = b.Concat(['x', 'rf', 'et'])
# layer that concatenates random forests outputs
layer_1 = cat(x=x, rf=rf(X=x, y=y), et=extra_trees(X=x, y=y))
# forest for the final prediction
final_extra_trees = b.ETC()
# pipeline output
b.Output('proba')(final_extra_trees(X=layer_1, y=y))
# build pipeline
pipeline = b.build()
# wrap pipeline into a scikit-learn model
model = BoskPipelineClassifier(pipeline, executor_cls=RecursiveExecutor)
# fit the model
model.fit(X_train, y_train)
# predict with the model
test_preds = model.predict(X_test)
For more examples look at Getting started