tfp.experimental.fastgp.partial_lanczos.psd_solve_multishift

Solve (A - shift_k I) x_k = v for PSD A.

tfp.experimental.fastgp.partial_lanczos.psd_solve_multishift(
    multiplier: Callable[[tfp.experimental.fastgp.fast_gp.Array], tfp.experimental.fastgp.fast_gp.Array],
    v: tfp.experimental.fastgp.fast_gp.Array,
    shifts: tfp.experimental.fastgp.fast_gp.Array,
    key: tfp.experimental.fastgp.fast_gp.Array,
    num_iters: int = 20
) -> tfp.experimental.fastgp.fast_gp.Array

multiplier A function for left matrix multiplying by an n by n symmetric positive definite matrix A.
v A tensor of shape (n, t) representing t n-dim vectors.
shifts A tensor of shape (s) representing s distinct scalar shifts.
key A random seed.
num_iters The number of iterations to run the Lanczos tridiagonalization algorithm for.

A tensor x of shape (s, t, n) that approximately satisfies (A - shift[i] I) x[i, j, :] = v[:, j]