squlearn.qnn.loss.ParameterRegularizationLoss

class squlearn.qnn.loss.ParameterRegularizationLoss(alpha: float | Callable[[int], float] = 0.005, mode: str = 'L2', parameter_list: list | None = None, parameter_operator_list: list | None = [])

Loss for parameter regularization.

Possible implementations:

  • "L1": \(L=\alpha \sum_i \left|p_i\right|\)

  • "L2": \(L=\alpha \sum_i p_i^2\)

Parameters:

  • alpha (float, Callable[[int], float]) – Weight value \(\alpha\)

  • mode (str) – Type of regularization, either ‘L1’ or ‘L2’ (default: ‘L2’).

  • parameter_list (list) – List of parameters to regularize, None: all (default: None).

  • parameter_operator_list (list) – List of operator parameters to regularize, None: all (default: []).

gradient(value_dict: dict, **kwargs) → ndarray | tuple[ndarray, ndarray]

Returns the gradient of the variance.

This function calculates the gradient of the variance values in value_dict.

Parameters:

  • value_dict (dict) – Contains calculated values of the model

  • iteration (int) – iteration number, if variance_factor is a function

Returns:

Gradient values

set_opt_param_op(opt_param_op: bool = True)

Sets the opt_param_op flag.

Parameters:

opt_param_op (bool) – True, if operator has trainable parameters

value(value_dict: dict, **kwargs) → float

Returns the variance.

This function returns the weighted variance as

\[L_\text{var} = \alpha \sum_i \var_i\]
Parameters:

  • value_dict (dict) – Contains calculated values of the model

  • iteration (int) – iteration number, if alpha is a callable function

Returns:

Loss value

variance(value_dict: dict, **kwargs) → float

Returns 0 since the variance is equal to zero.