A scikit-learn compatible implementation of Polynomial Spline/B-Spline as well as a Multidimensional/Tensor spline variation.
The BSpline class generates a one-dimensional spline for each feature. Knots are generated via histogram splits. The TensorBSplines gets a row-wise Kron-product between the splines until there is only one matrix left. Use with caution, as a few features will grow exponentially.
Parameters for both BSpline, TensorBSplines
n_bin: int
number of knots.
polynomial_degrees: int
polynomial degrees
sparse: bool
return sparse or dense matrix
pip install git+https://github.com/pr38/tensorbspline