propaq.noise

Noise models.

See the noise guide.

noise

propaq noise models.

Classes:

Name	Description
UniformNoiseModel	Exponential damping noise: each term of weight w is scaled by \(\exp(-\gamma w)\),
GateNoiseModel	A custom Python noise model.
NativeNoiseModel	Rust/C/AOT-compiled Julia noise model class.

UniformNoiseModel

Bases: UniformNoiseModel

Exponential damping noise: each term of weight w is scaled by \(\exp(-\gamma w)\), where \(w\) is the term's Pauli weight.

Parameters:

Name	Type	Description	Default
damping		Per-weight damping rate \(\gamma\). Each term is multiplied by \(\exp(-\gamma w)\).	required

Methods:

Name	Description
apply_noise	Apply uniform damping to all terms in term_sum in-place.
damping_factor	Return \(\exp(-\gamma w)\): the multiplicative factor applied to a term's coefficient.

Attributes:

Name	Type	Description
damping	float

damping property

damping: float

apply_noise method descriptor

apply_noise(term_sum: Any) -> None

Apply uniform damping to all terms in term_sum in-place.

Parameters:

Name	Type	Description	Default
term_sum	Any	A MajoranaTermSum or PauliTermSum to damp in-place.	required

damping_factor method descriptor

damping_factor(term_weight: int, active_modes: int) -> float

Return \(\exp(-\gamma w)\): the multiplicative factor applied to a term's coefficient.

Parameters:

Name	Type	Description	Default
term_weight	int	Pauli weight of the term.	required
active_modes	int	Unused for uniform noise; present for API compatibility.	required

GateNoiseModel

Bases: GateNoiseModel

A custom Python noise model.

Subclass this and define damping_factor or damping_factor_term directly. See NoiseModel for the two hook methods and the noise guide for worked examples of both.

NativeNoiseModel

Bases: NativeNoiseModel

Rust/C/AOT-compiled Julia noise model class.

A plugin declares what it reads through propaq_noise_depends.

• 0 is a function of term weight alone, so it is collapsed to one table indexed by weight before propagation starts and never called again.
• 2 (PROPAQ_DEPENDS_LAYER) also reads the circuit position. It keeps the tabulated fast path, but the table is rebuilt at each layer boundary.
• 1 (PROPAQ_DEPENDS_KEY) reads each term's raw basis-string words, necessary for structure-aware noise models.

The bits combine. See examples/plugins/README.md for the ABI and the example plugins.

Methods:

Name	Description
factor_term	Delegate to the plugin's propaq_noise_factor.

Attributes:

Name	Type	Description
abi_version	int	The ABI version the loaded plugin declared.
depends	int	The dependency bitmask the plugin declared: 1 = reads the term's key,

abi_version property

abi_version: int

The ABI version the loaded plugin declared.

depends property

depends: int

The dependency bitmask the plugin declared: 1 = reads the term's key, 2 = reads the layer index. 0 means a function of weight alone.

factor_term method descriptor

factor_term(basis_kind: int, words: Sequence[int], n_units: int, weight: int, layer_index: int = 0, n_layers: int = 0) -> float

Delegate to the plugin's propaq_noise_factor.

Exposed so a plugin can be exercised from Python without running a circuit; propagation calls the same entry point directly from the pool.

Parameters:

Name	Type	Description	Default
basis_kind	int	0 for Pauli, 1 for Majorana.	required
words	Sequence[int]	The term's raw basis-string words, two bits per unit. Ignored (and passed as NULL) unless the plugin declared it reads keys.	required
n_units	int	Qubits (Pauli) or modes (Majorana) of the register.	required
weight	int	The term's weight.	required
layer_index	int	Zero-based circuit layer.	0
n_layers	int	Layers in the circuit.	0

Base class

NoiseModel

Bases: ABC

Interface reference for a custom Python noise model.