Topic
Uncertainty, reliability and decisions with limits.
Two positive transport networks can lose exactly the same total mass, share every eigenvalue, and still disagree about a finite-time sensor alarm. A controlled study asks what a spectrum misses, what a rigorous margin can certify, and when that margin becomes uninformative.
A small error or an improved average answers only its own question. Event timing and worst-group reliability need their own checks.
A failed check can narrow a useful question. Two published numerical studies show why the original failure must remain visible.
A folding-energy study separates minima from saddles, global energy ordering from metastability, and branch discovery from proof, asking when an optimizer's attractive shape map leaves something important out.
Two spatial maps can share a convincing least-cost explanation without revealing the history that produced them. A controlled unbalanced optimal transport study separates endpoint fit, numerical convergence and evidence for movement versus local growth.
Delayed regional totals can support a precise-looking estimate without tightly constraining the hidden state. A six-region synthetic SIR benchmark compares ensemble intervals with validated set bounds, including the cost of bounds that become almost useless.
From a convincing fit to a useful guarantee: ask what the observations, evaluation and bounds actually establish.
First: When the Fit Is Not the Model
Start with a classroom timetable, then examine missing traffic states, numerical event errors and the limits of model reduction.
First: When the Timetable Lies to the Ventilation System
Sensitivity, calibration, robustness, and decision-making when model inputs or operating conditions are uncertain.
2026
A cohort-aware model of how slopes, stairs, transfers, crowding, and missing elevators change routes through a border hub.
2026
A verified stochastic population study compares fixed, proportional, and refuge-based harvesting near a nonlinear collapse threshold.
2026
A reproducible obstacle-PDE study resolves the American put exercise boundary and audits price, Greeks, convergence, and complementarity across three numerical methods.
2026
A calibrated graph-CUSUM benchmark for warning of local battery-pack thermal faults when a temperature sensor drifts or becomes unavailable.
2026
A study of verification-only entity repair separates finding the highest score from choosing a useful repair, and fewer complete evaluations from lower total cost.
2026
A reproducible cellular Potts study separates neighbour sorting from engulfment and tests adhesion, motility, initial conditions, energetics, and cell-shape plausibility.
