pdft.bases.BlockedBasis

class pdft.bases.BlockedBasis(inner, block_log_m, block_log_n, code=None, inv_code=None)

Bases: object

Wraps an inner parametric basis as a within-block transform.

Parameters:

• inner (any pdft basis (QFTBasis, EntangledQFTBasis, TEBDBasis, MERABasis)) – Within-block parametric circuit at smaller m_inner = inner.m, n_inner = inner.n.

• block_log_m (int) – Number of block-index qubits per dimension. The image is (2^(inner.m + block_log_m), 2^(inner.n + block_log_n)). block_log_m=0 (and =0) reduces to the inner basis.

• block_log_n (int) – Number of block-index qubits per dimension. The image is (2^(inner.m + block_log_m), 2^(inner.n + block_log_n)). block_log_m=0 (and =0) reduces to the inner basis.

• code (object)

• inv_code (object)

Notes

All learnable parameters live in inner.tensors; BlockedBasis is a pure structural wrapper. Block parameters are SHARED across blocks (one inner basis tiled).

__init__(inner, block_log_m, block_log_n, code=None, inv_code=None)

Parameters:

• inner (Any)

• block_log_m (int)

• block_log_n (int)

• code (object | None)

• inv_code (object | None)

Methods

__init__(inner, block_log_m, block_log_n[, ...])
forward_transform(pic)
inverse_transform(pic)

Attributes

block_shape
image_size
inv_tensors
m
n
num_blocks
num_parameters
tensors	Forward to inner — pytree leaves order is inner.tensors.
inner
block_log_m
block_log_n
code
inv_code

inner: Any

block_log_m: int

block_log_n: int

code: object

inv_code: object

property m: int

property n: int

property tensors: list[Array]

Forward to inner — pytree leaves order is inner.tensors.

property inv_tensors: list[Array]

property image_size: tuple[int, int]

property num_parameters: int

property num_blocks: int

property block_shape: tuple[int, int]

forward_transform(pic)

Parameters:

pic (Array)

Return type:

Array

inverse_transform(pic)

Parameters:

pic (Array)

Return type:

Array