Methodology
How DOMINO-Q computes – and what it does not claim
Dataset
- Bengaluru Industrial Corridor – Synthetic Demonstrator
- Seed v1.0: all probabilities, dependencies, impacts, interventions and effects are illustrative assumptions.
- Fictional asset locations around Bengaluru; not real infrastructure or a flood forecast.
- 10 assets, 12 directed dependencies, 12 candidate actions, 3 weighted scenarios.
Units
- Loss is measured in synthetic disruption points: impact weight × cascaded disruption probability, summed over assets.
- Disruption points are not rupees, lives or hours, and are not money saved.
- Costs are illustrative, expressed in ₹ lakh.
Model
- Local probability = base × scenario hazard multiplier × product of (1 − action effect), capped at 0.99.
- Three fixed propagation sweeps model conditional dependency; they are an iteration count, not a timeline.
- Risk measure: weighted CVaR at α = 0.8 (mean of the worst 20% probability mass of scenario losses).
Constraints
- Budget, maximum number of actions, lead time ≤ horizon, prerequisites (e.g. a11 requires a01) and mutual exclusions (a03 vs a10).
Solver
- Exact exhaustive enumeration on CPU of every portfolio; the reported optimum is provably optimal for this proxy model.
- Comparators: no action, a deliberately simple priority-first heuristic, and on request greedy and seeded simulated annealing.
What the headline % means
- The improvement vs priority-first is a difference in synthetic objective score against a naive baseline. It is not real risk reduction, savings, or quantum advantage. Where greedy or annealing match the optimum, no superiority over them is claimed.
What this is not
- Not a calibrated flood forecast or warning system.
- No live weather, satellite or sensor data is used.
- No quantum hardware is connected (Quantum hardware: NOT CONFIGURED).