series

Measurements on Pandas Series.

Classes

Aggregate	Aggregate a Pandas Series and produce a float or int.
NoisyQuantile	Estimates the quantile of a Pandas Series.
AddNoiseToSeries	A measurement that adds noise to each value in a pandas Series.

class Aggregate(input_domain, input_metric, output_measure, output_spark_type)

Bases: tmlt.core.measurements.base.Measurement

Aggregate a Pandas Series and produce a float or int.

Parameters

• input_domain (tmlt.core.domains.pandas_domains.PandasSeriesDomain) –

• input_metric (Union[tmlt.core.metrics.HammingDistance, tmlt.core.metrics.SymmetricDifference]) –

• output_measure (tmlt.core.measures.Measure) –

• output_spark_type (pyspark.sql.types.DataType) –

__init__(input_domain, input_metric, output_measure, output_spark_type)

Constructor.

Parameters

• input_domain (PandasSeriesDomainPandasSeriesDomain) – Input domain. Must have type PandasSeriesDomain.

• input_metric (HammingDistance | SymmetricDifferenceUnion[HammingDistance, SymmetricDifference]) – Input metric.

• output_measure (MeasureMeasure) – Output measure.

• output_spark_type (DataTypeDataType) – Spark DataType of the output. This is required to use this measurement within a udf.

property output_spark_type(self)

Return the Spark type of the aggregated value.

Return type

pyspark.sql.types.DataType

abstract __call__(self, data)

Perform measurement.

Parameters

data (pandas.Series) –

Return type

Union[float, int]

property input_domain(self)

Return input domain for the measurement.

Return type

tmlt.core.domains.base.Domain

property input_metric(self)

Distance metric on input domain.

Return type

tmlt.core.metrics.Metric

property output_measure(self)

Distance measure on output.

Return type

tmlt.core.measures.Measure

property is_interactive(self)

Returns true iff the measurement is interactive.

Return type

bool

privacy_function(self, d_in)

Returns the smallest d_out satisfied by the measurement.

See the privacy and stability tutorial (add link?) for more information.

Parameters

d_in (Any) – Distance between inputs under input_metric.

Raises

NotImplementedError – If not overridden.

Return type

Any

privacy_relation(self, d_in, d_out)

Return True if close inputs produce close outputs.

See the privacy and stability tutorial (add link?) for more information.

Parameters

• d_in (Any) – Distance between inputs under input_metric.

• d_out (Any) – Distance between outputs under output_measure.

Return type

bool

class NoisyQuantile(input_domain, output_measure, quantile, lower, upper, epsilon)

Bases: Aggregate

Estimates the quantile of a Pandas Series.

Methods

quantile()	Returns the quantile to be computed.
lower()	Returns the lower clamping bound.
upper()	Returns the upper clamping bound.
epsilon()	Returns the PureDP privacy budget to be used for producing a quantile.
privacy_function()	Returns the smallest d_out satisfied by the measurement.
__call__()	Return DP answer(float) to quantile query.
output_spark_type()	Return the Spark type of the aggregated value.
input_domain()	Return input domain for the measurement.
input_metric()	Distance metric on input domain.
output_measure()	Distance measure on output.
is_interactive()	Returns true iff the measurement is interactive.
privacy_relation()	Return True if close inputs produce close outputs.

Parameters

• input_domain (tmlt.core.domains.pandas_domains.PandasSeriesDomain) –

• output_measure (Union[tmlt.core.measures.PureDP, tmlt.core.measures.RhoZCDP]) –

• quantile (float) –

• lower (Union[float, int]) –

• upper (Union[float, int]) –

• epsilon (tmlt.core.utils.exact_number.ExactNumberInput) –

__init__(input_domain, output_measure, quantile, lower, upper, epsilon)

Constructor.

Parameters

• input_domain (PandasSeriesDomainPandasSeriesDomain) – Input domain. Must be PandasSeriesDomain.

• output_measure (PureDP | RhoZCDPUnion[PureDP, RhoZCDP]) – Output measure.

• quantile (floatfloat) – The quantile to produce.

• lower (float | intUnion[float, int]) – The lower clamping bound.

• upper (float | intUnion[float, int]) – The upper clamping bound.

• epsilon (ExactNumber | float | int | str | Fraction | ExprUnion[ExactNumber, float, int, str, Fraction, Expr]) – The pure-dp privacy parameter to use to produce the quantile.

property quantile(self)

Returns the quantile to be computed.

Return type

float

property lower(self)

Returns the lower clamping bound.

Return type

Union[float, int]

property upper(self)

Returns the upper clamping bound.

Return type

Union[float, int]

property epsilon(self)

Returns the PureDP privacy budget to be used for producing a quantile.

Return type

tmlt.core.utils.exact_number.ExactNumber

privacy_function(self, d_in)

Returns the smallest d_out satisfied by the measurement.

This algorithm uses the exponential mechanism, so benefits from the same privacy analysis:

If the output measure is PureDP, returns

\(\epsilon \cdot d_{in}\)

If the output measure is RhoZCDP, returns

\(\frac{1}{8}(\epsilon \cdot d_{in})^2\)

where:

• \(d_{in}\) is the input argument d_in

• \(\epsilon\) is epsilon

See [CR21] for the zCDP privacy analysis.

Parameters

d_in (tmlt.core.utils.exact_number.ExactNumberInput) – Distance between inputs under input_metric.

Return type

tmlt.core.utils.exact_number.ExactNumber

__call__(self, data)

Return DP answer(float) to quantile query.

TODO(#792) Add link to open-source paper: See this document for a description of the algorithm.

Parameters

data (pandas.Series) – The Series on which to compute the quantile.

Return type

float

property output_spark_type(self)

Return the Spark type of the aggregated value.

Return type

pyspark.sql.types.DataType

property input_domain(self)

Return input domain for the measurement.

Return type

tmlt.core.domains.base.Domain

property input_metric(self)

Distance metric on input domain.

Return type

tmlt.core.metrics.Metric

property output_measure(self)

Distance measure on output.

Return type

tmlt.core.measures.Measure

property is_interactive(self)

Returns true iff the measurement is interactive.

Return type

bool

privacy_relation(self, d_in, d_out)

Return True if close inputs produce close outputs.

See the privacy and stability tutorial (add link?) for more information.

Parameters

• d_in (Any) – Distance between inputs under input_metric.

• d_out (Any) – Distance between outputs under output_measure.

Return type

bool

class AddNoiseToSeries(noise_measurement)

Bases: tmlt.core.measurements.base.Measurement

A measurement that adds noise to each value in a pandas Series.

Parameters

noise_measurement (Union[tmlt.core.measurements.noise_mechanisms.AddLaplaceNoise, tmlt.core.measurements.noise_mechanisms.AddGeometricNoise, tmlt.core.measurements.noise_mechanisms.AddDiscreteGaussianNoise]) –

__init__(noise_measurement)

Constructor.

Parameters

noise_measurement (AddLaplaceNoise | AddGeometricNoise | AddDiscreteGaussianNoiseUnion[AddLaplaceNoise, AddGeometricNoise, AddDiscreteGaussianNoise]) – Noise Measurement to be applied to each element in input pandas Series.

property noise_measurement(self)

Returns measurement that adds noise to each number in pandas Series.

Return type

Union[tmlt.core.measurements.noise_mechanisms.AddLaplaceNoise, tmlt.core.measurements.noise_mechanisms.AddGeometricNoise, tmlt.core.measurements.noise_mechanisms.AddDiscreteGaussianNoise]

property input_domain(self)

Return input domain for the measurement.

Return type

tmlt.core.domains.pandas_domains.PandasSeriesDomain

property output_type(self)

Return the output data type after being used as a UDF.

Return type

pyspark.sql.types.DataType

privacy_function(self, d_in)

Returns the smallest d_out satisfied by the measurement.

Parameters

d_in (tmlt.core.utils.exact_number.ExactNumberInput) – Distance between inputs under input_metric.

Return type

tmlt.core.utils.exact_number.ExactNumber

__call__(self, values)

Adds noise to each number in the input Series.

Parameters

values (pandas.Series) –

Return type

pandas.Series

property input_metric(self)

Distance metric on input domain.

Return type

tmlt.core.metrics.Metric

property output_measure(self)

Distance measure on output.

Return type

tmlt.core.measures.Measure

property is_interactive(self)

Returns true iff the measurement is interactive.

Return type

bool

privacy_relation(self, d_in, d_out)

Return True if close inputs produce close outputs.

See the privacy and stability tutorial (add link?) for more information.

Parameters

• d_in (Any) – Distance between inputs under input_metric.

• d_out (Any) – Distance between outputs under output_measure.

Return type

bool