num_measurements_regularization.docstring

Return how many measurements we have from regularization

SYNOPSIS

    m = mrcal.cameramodel('xxx.cameramodel')

    optimization_inputs = m.optimization_inputs()

    x = mrcal.optimizer_callback(**optimization_inputs)[1]

    Nmeas   = mrcal.num_measurements_regularization( **optimization_inputs)
    i_meas0 = mrcal.measurement_index_regularization(**optimization_inputs)

    x_regularization = x[i_meas0:i_meas0+Nmeas]

The optimization algorithm tries to minimize the norm of a "measurements" vector
x. The optimizer doesn't know or care about the meaning of each element of this
vector, but for later analysis, it is useful to know what's what. The
mrcal.num_measurements_...() functions report where particular items end up in
the vector of measurements.

THIS function reports how many measurements come from the regularization terms
of the optimization problem. These don't model physical effects, but guide the
solver away from obviously-incorrect solutions, and resolve ambiguities. This
helps the solver converge to the right solution, quickly.

In order to determine the layout, we need quite a bit of context. If we have the
full set of inputs to the optimization function, we can pass in those (as shown
in the example above). Or we can pass the individual arguments that are needed
(see ARGUMENTS section for the full list). If the optimization inputs and
explicitly-given arguments conflict about the size of some array, the explicit
arguments take precedence. If any array size is not specified, it is assumed to
be 0. Thus most arguments are optional.

ARGUMENTS

- **kwargs: if the optimization inputs are available, they can be passed-in as
  kwargs. These inputs contain everything this function needs to operate. If we
  don't have these, then the rest of the variables will need to be given

- lensmodel: string specifying the lensmodel we're using (this is always
  'LENSMODEL_...'). The full list of valid models is returned by
  mrcal.supported_lensmodels(). This is required if we're not passing in the
  optimization inputs

- do_optimize_intrinsics_core
  do_optimize_intrinsics_distortions
  do_optimize_extrinsics
  do_optimize_frames
  do_optimize_calobject_warp
  do_apply_regularization

  optional booleans; default to True. These specify what we're optimizing. See
  the documentation for mrcal.optimize() for details

- Ncameras_intrinsics
  Ncameras_extrinsics
  Nframes
  Npoints
  Npoints_fixed
  Nobservations_board
  Nobservations_point
  calibration_object_width_n
  calibration_object_height_n

  optional integers; default to 0. These specify the sizes of various arrays in
  the optimization. See the documentation for mrcal.optimize() for details

RETURNED VALUE

The integer reporting how many elements of the measurement vector x come from
regularization terms