ResidualViolationScenario
Agents scenarios · Implementation: mlsysim/engine/scenario_definitions/agents.py.
Construct this scenario with keyword inputs. The fields below show the types and defaults declared by this class.
A residual per-step rate of violating proposals over horizon steps, for a day’s trajectories.
Notes
Equation: Trajectory violation probability is 1-(1-step_violation)**horizon; multiplying by daily trajectories gives an expected count.
Assumptions: Per-step violation events are independent with a fixed rate.
Excludes: Correlated violations, event severity, and distinction between proposed and executed harmful actions.
References
Derivation: the scenario composes mlsysim.physics.agents.calc_unchecked_success. Their contracts supply the underlying equations; property equations show the remaining unit conversions and accounting.
Examples
>>> from mlsysim import Scenarios
>>> scenario = Scenarios.ResidualViolationsFiftyTurns
>>> value = scenario.trajectory_violation
>>> round(float(value), 3)
0.049Attributes
| Name | Description |
|---|---|
| horizon | |
| step_violation | |
| trajectories_per_day | |
| trajectory_violation | Chance of at least one violating proposal under independent per-step events. |
| violating_per_day | Expected daily trajectories containing a violation, not a count of violation events. |
step_violation
Type: float.
Declared default: required.
horizon
Type: int.
Declared default: required.
trajectories_per_day
Type: int.
Declared default: required.
trajectory_violation
engine.scenario_types.ResidualViolationScenario.trajectory_violation
Chance of at least one violating proposal under independent per-step events.
Equation: 1 - calc_unchecked_success(1 - step_violation, horizon).
violating_per_day
engine.scenario_types.ResidualViolationScenario.violating_per_day
Expected daily trajectories containing a violation, not a count of violation events.
Equation: trajectory_violation * trajectories_per_day.