pdft.optimizers.RiemannianGD

class pdft.optimizers.RiemannianGD(lr=0.01, armijo_c=0.0001, armijo_tau=0.5, max_ls_steps=10, max_grad_norm=None)

Bases: object

Riemannian gradient descent with Armijo backtracking line search.

Mirror of upstream src/optimizers.jl:156-166. Defaults match upstream.

Parameters:

• lr (float)

• armijo_c (float)

• armijo_tau (float)

• max_ls_steps (int)

• max_grad_norm (float | None)

__init__(lr=0.01, armijo_c=0.0001, armijo_tau=0.5, max_ls_steps=10, max_grad_norm=None)

Parameters:

• lr (float)

• armijo_c (float)

• armijo_tau (float)

• max_ls_steps (int)

• max_grad_norm (float | None)

Return type:

None

Methods

__init__([lr, armijo_c, armijo_tau, ...])

Attributes

armijo_c
armijo_tau
lr
max_grad_norm
max_ls_steps

lr: float = 0.01

armijo_c: float = 0.0001

armijo_tau: float = 0.5

max_ls_steps: int = 10

max_grad_norm: float | None = None