quara.loss_function.loss_function module

class LossFunction(num_var=None)[source]

Bases: object

Constructor

Subclasses have a responsibility to set on_value, on_gradient, on_hessian.

Parameters: num_var (int, optional) – number of variables, by default None

abstract gradient(var)[source]

returns the gradient of the loss function.

Parameters: var (np.ndarray) – np.ndarray of variables.
Returns: the gradient of the loss function. dtype=np.float64
Return type: np.ndarray
Raises: NotImplementedError – this function does not be implemented in the subclass.

abstract hessian(var)[source]

returns the Hessian of the loss function.

Parameters: var (np.ndarray) – np.ndarray of variables.
Returns: the Hessian of the loss function. dtype=np.float64
Return type: np.ndarray
Raises: NotImplementedError – this function does not be implemented in the subclass.

is_option_sufficient()[source]

returns whether the option is sufficient.

In the default implementation, this function returns True. Override with subclasses as needed.

Returns: whether the option is sufficient.
Return type: bool

property num_var: int

returns number of variables.

Returns: number of variables.
Return type: int

property on_gradient: bool

returns whether or not to support gradient.

Returns: whether or not to support gradient.
Return type: bool

property on_hessian: bool

returns whether or not to support Hessian.

Returns: whether or not to support Hessian.
Return type: bool

property on_value: bool

returns whether or not to support value.

Returns: whether or not to support value.
Return type: bool

property option: quara.loss_function.loss_function.LossFunctionOption

returns loss function option.

Returns: loss function option.
Return type: LossFunctionOption

set_from_option(option)[source]

sets option from LossFunctionOption and calls is_option_sufficient function.

Parameters: option (LossFunctionOption) – option to set.
Return type: None

set_from_standard_qtomography_option_data(qtomography, option, data, is_gradient_required, is_hessian_required)[source]

sets settings of loss function.

This implementation is empty by default. If necessary, implement it in subclasses.

Parameters

qtomography (StandardQTomography) – StandardQTomography for settings of loss function.
option (LossFunctionOption) – ProbabilityBasedLossFunctionOption for settings of loss function.
data (List[Tuple[int, np.ndarray]]) – empirical distributions for settings of loss function.
is_gradient_required (bool) – whether or not to require gradient.
is_hessian_required (bool) – whether or not to require Hessian.

Return type

None

abstract value(var)[source]

returns the value of the loss function.

Parameters: var (np.ndarray) – np.ndarray of variables.
Returns: the value of the loss function.
Return type: np.float64
Raises: NotImplementedError – this function does not be implemented in the subclass.

class LossFunctionOption(mode_weight=None, weights=None, weight_name=None)[source]

Bases: object

Constructor

Parameters

mode_weight (str, optional) – mode weight, by default None
weights (List, optional) – list of weight, by default None
weight_name (str, optional) – weight name for reporting, by default None

property mode_weight: str

returns mode weight.

Returns: mode weight.
Return type: str

property weight_name: str

returns weight name for reporting.

Returns: weight name for reporting.
Return type: str

property weights: List

returns weights.

Returns: list of weight.
Return type: List