DCRABOptimizerGradient

class DCRABOptimizerGradient(measure_and_gradient_func, optimization_map, super_iteration_every=30, max_super_iteration_num=10, print_every_iteration_num=5, fallback_optimization=None, super_iteration_tol=1e-07, seed=None)

Bases: ScipyOptimizerGradient

A dCRAB optimization method [3, 14].

Implements dCRAB optimization involving super-iterations that add additional optimization components to the dCRAB envelope and freezes the older parameters.

Note

This works with a DCRABEnvelope or a list[DCRABEnvelope] as envelopes.

__init__(measure_and_gradient_func, optimization_map, super_iteration_every=30, max_super_iteration_num=10, print_every_iteration_num=5, fallback_optimization=None, super_iteration_tol=1e-07, seed=None)

Parameters:

• measure_and_gradient_func (Callable[[Array], tuple[Float, Array]]) – Function implementing measurement of observables to be minimized, returning (value, gradient).

• optimization_map (OptimizationMap) – An optimization map containing all parameters that can be optimized.

• super_iteration_every (int) – Number of iterations after which one super-iteration is performed. Defaults to 30.

• max_super_iteration_num (int) – Maximum number of super-iterations to perform. Defaults to 10.

• print_every_iteration_num (int) – Print every this many iterations the current optimization value. Defaults to 5.

• fallback_optimization (Optimizer | None) – Fallback optimization to perform after all super-iterations.

• super_iteration_tol (float) – Tolerance for triggering a super-iteration. Defaults to 1e-7.

• seed (int | None) – Random seed for adding new dCRAB components.

Methods

__init__(measure_and_gradient_func, ...[, ...])
optimize(times)	Optimize via the Scipy optimizer gradient model.
set_options(opts)	Set the options for the system.
set_parameters(values)	Update the parameter values.
update_option(key, val)	Update one option for the system.

Attributes

best_fid
callback	Returns the callback function.
logger	Returns the current logger that is being used by this optimizer, or None if no logger was set yet.
method	Returns the currently selected optimization method.
optimization_map	Return the optimization map that this optimizer uses.
best_params

property callback: Callable | None

Returns the callback function.

property logger: Logger | None

Returns the current logger that is being used by this optimizer, or None if no logger was set yet.

property method: str

Returns the currently selected optimization method.

property optimization_map: OptimizationMap

Return the optimization map that this optimizer uses.

Parameters that can be optimized need to be added to this map.

Returns:

The optimization map that this optimizer uses.

optimize(times)

Optimize via the Scipy optimizer gradient model.

Performs the actual optimization.

Note

If input times is a float, then the start time of propagation is implicitly assumed to be zero. For an array of times, the first time point is the start time.

Parameters:

times (Array | float) – Array of times or a float (assumed start time zero).

Returns:

The result of the optimization.

Return type:

OptimizationResult

set_options(opts)

Set the options for the system.

set_parameters(values)

Update the parameter values.

This method is derived from the ScipyOptimizerGradient class and designed to catch cases involving mismatch in dimension of parameters.

Since, in dCRAB, new parameters are added in each super-iteration, the previous best result may be one with fewer parameters. In that case, all subsequent parameters are set to their minimum value (by setting the reduced value to -1).

Parameters:

values (Array | list[float]) – Parameter values for the update.

Returns:

The log of parameter updates.

Return type:

list[Quantity]

update_option(key, val)

Update one option for the system.