Algorithms#

SenSR: Sensitive Subspace Robustness#

class inFairness.fairalgo.SenSR(network, distance_x, loss_fn, eps, lr_lamb, lr_param, auditor_nsteps, auditor_lr)[source]#

Implementes the Sensitive Subspace Robustness (SenSR) algorithm.

Proposed in Training individually fair ML models with sensitive subspace robustness

Parameters:

network (torch.nn.Module) – Network architecture
distance_x (inFairness.distances.Distance) – Distance metric in the input space
loss_fn (torch.nn.Module) – Loss function
eps (float) – \(\epsilon\) parameter in the SenSR algorithm
lr_lamb (float) – \(\lambda\) parameter in the SenSR algorithm
lr_param (float) – \(\alpha\) parameter in the SenSR algorithm
auditor_nsteps (int) – Number of update steps for the auditor to find worst-case examples
auditor_lr (float) – Learning rate for the auditor

forward(X, Y=None, *args, **kwargs)[source]#

Defines the computation performed at every call.

Parameters:

X (torch.Tensor) – Input data
Y (torch.Tensor) – Expected output data

Returns:

output – Model output

Return type:

torch.Tensor

forward_test(X)[source]#: Forward method during the test phase

forward_train(X, Y)[source]#: Forward method during the training phase

SenSeI: Sensitive Set Invariance#

class inFairness.fairalgo.SenSeI(network, distance_x, distance_y, loss_fn, rho, eps, auditor_nsteps, auditor_lr)[source]#

Implementes the Sensitive Set Invariane (SenSeI) algorithm.

Proposed in SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness

Parameters:

network (torch.nn.Module) – Network architecture
distance_x (inFairness.distances.Distance) – Distance metric in the input space
distance_y (inFairness.distances.Distance) – Distance metric in the output space
loss_fn (torch.nn.Module) – Loss function
rho (float) – \(\rho\) parameter in the SenSR algorithm
eps (float) – \(\epsilon\) parameter in the SenSR algorithm
auditor_nsteps (int) – Number of update steps for the auditor to find worst-case examples
auditor_lr (float) – Learning rate for the auditor

forward(X, Y=None, *args, **kwargs)[source]#

Defines the computation performed at every call.

Parameters:

X (torch.Tensor) – Input data
Y (torch.Tensor) – Expected output data

Returns:

output – Model output

Return type:

torch.Tensor

forward_test(X)[source]#: Forward method during the test phase

forward_train(X, Y)[source]#: Forward method during the training phase

SenSTIR: Sensitive Set Transport Invariant Ranking#

class inFairness.fairalgo.SenSTIR(network: Module, distance_x: MahalanobisDistances, distance_y: MahalanobisDistances, rho: float, eps: float, auditor_nsteps: int, auditor_lr: float, monte_carlo_samples_ndcg: int)[source]#

Implementes the Sensitive Subspace Robustness (SenSR) algorithm.

Proposed in Individually Fair Ranking

Parameters:

network (torch.nn.Module) – Network architecture
distance_x (inFairness.distances.Distance) – Distance metric in the input space
distance_y (inFairness.distances.Distance) – Distance metric in the output space
rho (float) – \(\rho\) parameter in the SenSTIR algorithm (see Algorithm 1)
eps (float) – \(\epsilon\) parameter in the SenSTIR algorithm (see Algorithm 1)
auditor_nsteps (int) – Number of update steps for the auditor to find worst-case examples
auditor_lr (float) – Learning rate for the auditor
monte_carlo_samples_ndcg (int) – Number of monte carlo samples required to estimate the gradient of the empirical version of expectation defined in equation SenSTIR in the reference

forward(Q, relevances, **kwargs)[source]#

Defines the computation performed at every call.

Parameters:

X (torch.Tensor) – Input data
Y (torch.Tensor) – Expected output data

Returns:

output – Model output

Return type:

torch.Tensor

forward_test(Q)[source]#: Forward method during the test phase