Amber provides code to evaluate multivariate Bernstein polynomials in TensorFlow using de Casteljau’s algorithm. Multivariate Bernstein basis polynomials are particularly useful for linear multivariate regression with linear inequality constraints.

Getting started

Amber requires miniconda. To test the Amber software cd into the project root directory and type:

conda env create --file=./environment.yml
conda activate amber
pytest

Further reading

Mama Foupouagnigni, Merlin Mouafo Wouodjié (2020). “On Multivariate Bernstein Polynomials” Mathematics 8, 1397. https://doi.org/10.3390/math8091397

Esmeralda Mainar, J.M. Pena (2006). “Evaluation algorithms for multivariate polynomials in Bernstein-Bezier form.” Journal of Approximation Theory 143, 44-61. https://doi.org/10.1016/j.jat.2006.05.007.