BiasFeedbackScenario

Statistics scenarios · Implementation: mlsysim/engine/scenario_definitions/statistics.py.

Construct this scenario with keyword inputs. The fields below show the types and defaults declared by this class.

A gap passed at launch, refit from the next audit, iterated over retraining cycles: gap(k) = launch_gap * factor^k against a release gate (all in percentage points).

Notes

Equation: Disparity compounds as launch_gap*factor**cycles; factor is fitted from the next audit and the gate-crossing index is rounded upward.

Assumptions: One constant multiplicative factor represents each retraining cycle; gap and gate use matching percentage-point units.

Excludes: A causal feedback model, saturation, sampling uncertainty, and assurance that the fitted factor persists.

References

Derivation: the scenario composes mlsysim.physics.fairness.calc_cycles_to_gate, mlsysim.physics.fairness.calc_feedback_disparity. Their contracts supply the underlying equations; property equations show the remaining unit conversions and accounting.

Examples

>>> from mlsysim import Scenarios
>>> scenario = Scenarios.CreditServiceBiasFeedback
>>> value = scenario.factor
>>> round(float(value), 3)
1.1

Attributes

Name	Description
breach_cycle	Ceiling of the continuous gate-crossing index, including equality with the gate.
crossing	Continuous cycle index at which geometric disparity reaches the gate.
factor	Multiplicative per-cycle disparity factor fitted from the first two audits.
gap_at_breach	Fitted disparity at the integer gate-crossing cycle, in percentage points.
gate
horizon_cycles
launch_gap
next_audit_gap
sampled_factor
sampled_gap_at_breach	Gap at the original crossing cycle under the alternative sampled factor.
unmitigated_at_horizon	Unmitigated launch gap compounded through the declared horizon, in percentage points.
unmitigated_gap

Methods

Name	Description
gap_at	Disparity in percentage points after `cycles` under the chosen multiplicative factor.

gap_at

engine.scenario_types.BiasFeedbackScenario.gap_at(cycles, factor=None)

Disparity in percentage points after cycles under the chosen multiplicative factor.

Equation: calc_feedback_disparity(launch_gap, factor or factor, cycles).