pylops.optimization.callback.Callbacks¶

class pylops.optimization.callback.Callbacks[source]¶

This is a template class which a user must subclass when implementing callbacks for a solver. This class comprises of the following methods:

on_setup_begin: a method that is invoked at the start of the setup method of the solver
on_setup_end: a method that is invoked at the end of the setup method of the solver
on_step_begin: a method that is invoked at the start of the step method of the solver
on_step_end: a method that is invoked at the end of the setup step of the solver
on_run_begin: a method that is invoked at the start of the run method of the solver
on_run_end: a method that is invoked at the end of the run method of the solver

All methods take two input parameters: the solver itself, and the vector x.

Moreover, some callback may be used to implement custom stopping criteria for the solver. This can be done by adding a boolean attribute stop to the callback object, which will be initially set to False. As soon as the callback sets this attribute to True, the run method of the solver will stop iterating and return the current model vector.

Examples

>>> import numpy as np
>>> from pylops.basicoperators import MatrixMult
>>> from pylops.optimization.basic import CG
>>> from pylops.optimization.callback import Callbacks
>>>
>>> class StoreIterCallback(Callbacks):
...     def __init__(self):
...         self.stored = []
...     def on_step_end(self, solver, x):
...         self.stored.append(solver.iiter)
>>>
>>> Aop = MatrixMult(np.random.normal(0., 1., 36).reshape(6, 6))
>>> Aop = Aop.H @ Aop
>>> y = Aop @ np.ones(6)
>>> cb_sto = StoreIterCallback()
>>> cgsolve = CG(Aop, callbacks=[cb_sto, ])
>>> xest = cgsolve.solve(y=y, x0=np.zeros(6), tol=0, niter=6, show=False)[0]
>>> xest, cb_sto.stored
(array([1., 1., 1., 1., 1., 1.]), [1, 2, 3, 4, 5, 6])

Methods

`__init__`()
`on_run_begin`(solver, x)	Callback before entire solver run
`on_run_end`(solver, x)	Callback after entire solver run
`on_setup_begin`(solver, x0)	Callback before setup
`on_setup_end`(solver, x)	Callback after setup
`on_step_begin`(solver, x)	Callback before step of solver
`on_step_end`(solver, x)	Callback after step of solver

Callbacks.on_run_begin(solver, x)[source]¶

Callback before entire solver run

Parameters:

solverpylops.optimization.basesolver.Solver: Solver object
xnumpy.ndarray: Current model vector

Callbacks.on_run_end(solver, x)[source]¶

Callback after entire solver run

Parameters:

solverpylops.optimization.basesolver.Solver: Solver object
xnumpy.ndarray: Current model vector

Callbacks.on_setup_begin(solver, x0)[source]¶

Callback before setup

Parameters:

solverpylops.optimization.basesolver.Solver: Solver object
x0numpy.ndarray: Initial guess (when present as one of the inputs of the solver setup method)

Callbacks.on_setup_end(solver, x)[source]¶

Callback after setup

Parameters: