Scatterplot Smoothers Documentation

This repository provides minimal and efficient implementations of key scatterplot smoothers, specifically Smoothing Splines and LOESS.

The core logic is implemented in C++ with Python bindings provided by pybind11.

Key Smoothers¶

• Smoothing Splines: Similar to smooth.spline in R, offering multiple fitting engines:

  • Reinsch Algorithm: $O(N)$ performance for when knots equal data points (matches R’s smooth.spline).

  • Natural Spline Basis: Explicit basis construction, suitable for regression splines ($K < N$).

  • B-Spline Basis: Efficient banded solver implementation using LAPACK.

• LOESS: A fast C++ implementation of locally estimated scatterplot smoothing (local polynomial regression).

See the table of contents for more details on the theory and comparisons with other implementations.