pydvl.influence.base_influence_function_model

InfluenceMode

Bases: str, Enum

Enum representation for the types of influence.

ATTRIBUTE	DESCRIPTION
Up	Approximating the influence of a point
Perturbation	Perturbation definition of the influence score

InfluenceFunctionModel

Bases: Generic[TensorType, DataLoaderType], ABC

Generic abstract base class for computing influence related quantities. For a specific influence algorithm and tensor framework, inherit from this base class

n_parameters abstractmethod property

n_parameters

Number of trainable parameters of the underlying model

is_thread_safe abstractmethod property

is_thread_safe: bool

Whether the influence computation is thread safe

is_fitted abstractmethod property

is_fitted

Override this, to expose the fitting status of the instance.

fit abstractmethod

fit(data: DataLoaderType) -> InfluenceFunctionModel

Override this method to fit the influence function model to training data, e.g. pre-compute hessian matrix or matrix decompositions

PARAMETER	DESCRIPTION
data	TYPE: DataLoaderType

RETURNS	DESCRIPTION
InfluenceFunctionModel	The fitted instance

Source code in src/pydvl/influence/base_influence_function_model.py

@abstractmethod
def fit(self, data: DataLoaderType) -> InfluenceFunctionModel:
    """
    Override this method to fit the influence function model to training data,
    e.g. pre-compute hessian matrix or matrix decompositions

    Args:
        data:

    Returns:
        The fitted instance
    """

influences_from_factors abstractmethod

influences_from_factors(
    z_test_factors: TensorType,
    x: TensorType,
    y: TensorType,
    mode: InfluenceMode = InfluenceMode.Up,
) -> TensorType

Override this method to implement the computation of

\[ \langle z_{\text{test_factors}}, \nabla_{\theta} \ell(y, f_{\theta}(x)) \rangle \]

for the case of up-weighting influence, resp.

\[ \langle z_{\text{test_factors}}, \nabla_{x} \nabla_{\theta} \ell(y, f_{\theta}(x)) \rangle \]

for the perturbation type influence case. The gradient is meant to be per sample of the batch \((x, y)\).

PARAMETER	DESCRIPTION
z_test_factors	pre-computed array, approximating \(H^{-1}\nabla_{\theta} \ell(y_{\text{test}}, f_{\theta}(x_{\text{test}}))\) TYPE: TensorType
x	model input to use in the gradient computations \(\nabla_{\theta}\ell(y, f_{\theta}(x))\), resp. \(\nabla_{x}\nabla_{\theta}\ell(y, f_{\theta}(x))\), if None, use \(x=x_{\text{test}}\) TYPE: TensorType
y	label tensor to compute gradients TYPE: TensorType
mode	enum value of InfluenceMode TYPE: InfluenceMode DEFAULT: Up

RETURNS	DESCRIPTION
TensorType	Tensor representing the element-wise scalar products for the provided batch

Source code in src/pydvl/influence/base_influence_function_model.py

@abstractmethod
def influences_from_factors(
    self,
    z_test_factors: TensorType,
    x: TensorType,
    y: TensorType,
    mode: InfluenceMode = InfluenceMode.Up,
) -> TensorType:
    r"""
    Override this method to implement the computation of

    \[ \langle z_{\text{test_factors}},
        \nabla_{\theta} \ell(y, f_{\theta}(x)) \rangle \]

    for the case of up-weighting influence, resp.

    \[ \langle z_{\text{test_factors}},
        \nabla_{x} \nabla_{\theta} \ell(y, f_{\theta}(x)) \rangle \]

    for the perturbation type influence case. The gradient is meant to be per sample
    of the batch $(x, y)$.

    Args:
        z_test_factors: pre-computed array, approximating
            $H^{-1}\nabla_{\theta} \ell(y_{\text{test}},
            f_{\theta}(x_{\text{test}}))$
        x: model input to use in the gradient computations
            $\nabla_{\theta}\ell(y, f_{\theta}(x))$,
            resp. $\nabla_{x}\nabla_{\theta}\ell(y, f_{\theta}(x))$,
            if None, use $x=x_{\text{test}}$
        y: label tensor to compute gradients
        mode: enum value of [InfluenceMode]
            [pydvl.influence.base_influence_function_model.InfluenceMode]

    Returns:
        Tensor representing the element-wise scalar products for the provided batch

    """