parameters.py

Source: LION/reconstructors/diffusion/padis/parameters.py

Parameter presets and validation for the PaDIS reconstructor.

class LION.reconstructors.diffusion.padis.parameters.PaDISParameters

Bases: object

Define PaDIS parameter presets, merging, and validation.

static default_parameters(model=None)

Return conservative generic sampler parameters.

Study-specific reconstruction should normally start from one of the CT preset constructors rather than these diagnostic defaults.

Parameters:

model (NCSNpp | None)

Return type:

LIONParameter

static paper_ct_parameters(model=None, *, views=20)

Return the strict CT reconstruction sampler described in the paper.

Parameters:

• model (NCSNpp | None)

• views (int)

Return type:

LIONParameter

static lion_physics_ct_parameters(model=None, *, views=20)

Return the LION-native CT sampler with physical operator scaling.

The paper sigma schedule is kept unchanged, but the measurement update is expressed in LION CT units: residual gradients use the LION forward operator and adjoint, normalized by the composed measurement operator Lipschitz constant rather than by public-repository matching constants.

Parameters:

• model (NCSNpp | None)

• views (int)

Return type:

LIONParameter

static padis_repo_ct_parameters(model=None)

Return the public PaDIS CT-script sampler settings for compatibility.

Parameters:

model (NCSNpp | None)

Return type:

LIONParameter

static lion_quality_ct_parameters(model=None, *, views=20)

Return the preferred LION CT sampler for PaDIS reconstruction checks.

This keeps the paper’s CT sigma schedule and squared-residual data objective, but uses FDK initialization and operator-norm measurement scaling for stable LION-native CT operations.

Parameters:

• model (NCSNpp | None)

• views (int)

Return type:

LIONParameter