GaussianAccountant¶

class opacus.accountants.gdp.GaussianAccountant[source]¶

get_epsilon(delta, poisson=True)[source]¶

Return privacy budget (epsilon) expended so far.

Parameters:

• delta (float) – target delta

• poisson (bool) – True is input batches was sampled via Poisson sampling, False otherwise

Return type:

float

classmethod mechanism()[source]¶

Accounting mechanism name

Return type:

str

step(*, noise_multiplier, sample_rate)[source]¶

Signal one optimization step

Parameters:

• noise_multiplier (float) – Current noise multiplier

• sample_rate (float) – Current sample rate