Skip to content

cagpjax.solvers.cholesky

Linear solvers based on Cholesky decomposition.

Classes:

  • Cholesky –

    Solve a linear system using the Cholesky decomposition.

Cholesky 

Cholesky(jitter: ScalarFloat | None = None)

Bases: AbstractLinearSolver[CholeskyState]

Solve a linear system using the Cholesky decomposition.

Due to numerical imprecision, Cholesky factorization may fail even for positive-definite \(A\). Optionally, a small amount of jitter (\(\epsilon\)) can be added to \(A\) to ensure positive-definiteness. Note that the resulting system solved is slightly different from the original system.

Attributes:

  • jitter (ScalarFloat | None) –

    Small amount of jitter to add to \(A\) to ensure positive-definiteness.

Source code in src/cagpjax/solvers/cholesky.py 
31
32
def __init__(self, jitter: ScalarFloat | None = None):
    self.jitter = jitter