References

cholrot implements standard rank-one Cholesky update and downdate algorithms and exposes them through a compact NumPy-style API. The package does not claim novelty for the classical rotation methods themselves. Its practical focus is on making these routines easy to install, test, benchmark, and use from Python, with direct APIs for modified-factor products and rank-one modified solves.

Background references

  • LINPACK Cholesky update and downdate routines, including DCHUD and DCHDD.

  • Seeger, M. Low Rank Updates for the Cholesky Decomposition.

  • Chambers-style hyperbolic rotations for Cholesky downdating.

Ecosystem comparisons

The closest public interfaces are classical cholupdate-style routines. The main difference in cholrot is that the package also exposes product and solve operations that do not require materializing the modified factor.

Useful comparison points include:

  • MATLAB cholupdate;

  • JAX jax.lax.linalg.cholesky_update;

  • TensorFlow Probability tfp.math.cholesky_update.