aquakin.SensitivityResult#

class aquakin.SensitivityResult(output, doutput_dparams, doutput_dconditions, parameter_names)[source]#

Bases: object

Gradients of a scalar output with respect to parameters and conditions.

Variables:

  • output (float) – The scalar output value at the evaluation point.

  • doutput_dparams (jnp.ndarray) – Gradient w.r.t. the full flat params vector, shape (n_params,) — every model parameter, not a free subset (unlike fit() / calibrate(), which optimise a chosen free_params list).

  • doutput_dconditions (dict[str, jnp.ndarray]) – Gradient w.r.t. each condition field, field_name -> (n_locations,).

  • parameter_names (list[str]) – Namespaced parameter names matching doutput_dparams (all of them).

Parameters:

  • output (float)

  • doutput_dparams (Array)

  • doutput_dconditions (dict[str, Array])

  • parameter_names (list[str])

__init__(output, doutput_dparams, doutput_dconditions, parameter_names)#
Parameters:

  • output (float)

  • doutput_dparams (Array)

  • doutput_dconditions (dict[str, Array])

  • parameter_names (list[str])

Return type:

None

Methods

__init__(output, doutput_dparams, ...)

ranked_params()

Return (name, |grad|) pairs sorted by decreasing magnitude.

Attributes

output

doutput_dparams

doutput_dconditions

parameter_names

ranked_params()[source]#

Return (name, |grad|) pairs sorted by decreasing magnitude.

Return type:

list[tuple[str, float]]