solvers.ResponsibleEngineeringModel

solvers.ResponsibleEngineeringModel()

Models the computational cost of responsible AI practices (Wall 20: Safety).

This model quantifies the ‘Safety Tax’ — the additional compute and data required for differential privacy or fairness guarantees.

Literature Source: 1. Abadi et al. (2016), “Deep Learning with Differential Privacy.” 2. Anil et al. (2022), “Large-Scale Differentially Private BERT.”

Methods

Name	Description
solve	Calculates the overhead of responsible engineering practices.

solve

solvers.ResponsibleEngineeringModel.solve(
    base_training_time,
    epsilon=1.0,
    delta=1e-05,
    min_subgroup_prevalence=0.01,
)

Calculates the overhead of responsible engineering practices.