tfp.experimental.fastgp.GaussianProcessConfig

Configuration for distributions in the FastGP family.

tfp.experimental.fastgp.GaussianProcessConfig(
    cg_iters: int = 25,
    preconditioner: str = 'auto',
    preconditioner_rank: int = 25,
    preconditioner_num_iters: int = 5,
    precondition_before_jitter: str = 'auto',
    probe_vector_type: str = 'rademacher',
    num_probe_vectors: int = 35,
    log_det_algorithm: str = 'r3',
    log_det_iters: int = 20
)

cg_iters Dataclass field
preconditioner Dataclass field
preconditioner_rank Dataclass field
preconditioner_num_iters Dataclass field
precondition_before_jitter Dataclass field
probe_vector_type Dataclass field
num_probe_vectors Dataclass field
log_det_algorithm Dataclass field
log_det_iters Dataclass field

Methods

__eq__

__eq__(
    other
)

Return self==value.

cg_iters 25
log_det_algorithm 'r3'
log_det_iters 20
num_probe_vectors 35
precondition_before_jitter 'auto'
preconditioner 'auto'
preconditioner_num_iters 5
preconditioner_rank 25
probe_vector_type 'rademacher'