smash.Optimize

class smash.Optimize(data=None)

Represents optimize optional results.

See also

smash.optimize

Model assimilation using numerical optimization algorithms.

Notes

The object’s available attributes depend on what is requested by the user during a call to optimize in return_options.

Attributes:

time_steppandas.DatetimeIndex

A list of length n containing the returned time steps.

rr_statesFortranDerivedTypeArray

A list of length n of RR_StatesDT for each time_step.

q_domainnumpy.ndarray

An array of shape (nrow, ncol, n) representing simulated discharges on the domain for each time_step.

iter_costnumpy.ndarray

An array of shape (m,) representing cost iteration values from m iterations.

iter_projgnumpy.ndarray

An array of shape (m,) representing infinity norm of the projected gardient iteration values from m iterations.

control_vectornumpy.ndarray

An array of shape (k,) representing the control vector at end of optimization.

netNet

The trained neural network.

costfloat

Cost value.

jobsfloat

Cost observation component value.

jregfloat

Cost regularization component value.

lcurve_wjregdict[str, Any]

A dictionary containing the wjreg lcurve data. The elements are:

wjreg_optfloat

The optimal wjreg value.

distancenumpy.ndarray

An array of shape (6,) representing the L-Curve distance for each optimization cycle (the maximum distance corresponds to the optimal wjreg).

costnumpy.ndarray

An array of shape (6,) representing the cost values for each optimization cycle.

jobsnumpy.ndarray

An array of shape (6,) representing the jobs values for each optimization cycle.

jregnumpy.ndarray

An array of shape (6,) representing the jreg values for each optimization cycle.

wjregnumpy.ndarray

An array of shape (6,) representing the wjreg values for each optimization cycle.