Gaussian process solvers improve linear PDE solutions with inequality constraints

Constrained Probabilistic Numerics: Gaussian Process Solvers for Linear PDEs under Linear Inequality Constraints

Machine Learning

Summary

Solving mathematical equations called linear PDEs often gives uncertain results that satisfy exact conditions but ignore important limits like staying nonnegative or monotone. The authors show how to add such inequality limits into Gaussian process solvers without changing the basic approach. This leads to more accurate and physically realistic solutions in examples like image reconstruction, financial option pricing, and air quality prediction. Their method turns checking these limits from a separate step into a core part of the solution process.

What this means in practice

  • For quantitative finance teams: Enforce no-arbitrage constraints in option price modeling to reduce prediction errors and arbitrage violations with limited market data.
  • For environmental monitoring teams: Improve air-quality field reconstructions from sparse sensor data by incorporating nonnegativity constraints into PDE-based probabilistic models.

Authors

Boyuan Deng, Kshitiz Upadhyay, Michael Shields

Abstract

Gaussian process solvers for linear partial differential equations (PDEs) return posteriors rather than point estimates, but they condition only on equality information: the PDE residual, boundary data, and measurements. Many solutions of interest also obey inequality constraints, including nonnegativity, monotonicity, and absence of arbitrage, which equality conditioning cannot enforce. We prove that conditioning a Gaussian process prior jointly on linear equality observations and a finite-dimensional inequality event is equivalent to conditioning sequentially in terms of Borel measures on the solution space. The physics-informed posterior is therefore a valid prior for the inequality update at virtual points. The combined law is non-Gaussian yet computable through truncated-Gaussian moments, and neither the PDE conditioning nor the inequality update requires modification. We evaluated the construction on a Poisson benchmark, option-price surfaces, and air-quality fields. Under incomplete boundary data, a monotonicity constraint lowered the error of a two-dimensional Poisson reconstruction by $63\%$. On a real S&P 500 option chain thinned to $20$ retained quotes, no-arbitrage constraints reduced held-out extrapolation error by $22\%$ and cut the calendar-arbitrage rate from $10\%$ to $1\%$. On four real air-quality snapshots, nonnegativity reduced a Gaussian moment-based sub-zero-tail proxy by $74\%$ at the sparsest monitor count. Changes in mean accuracy and constraint diagnostics differed across data regimes. The probability of the noisy virtual constraint event under the equality posterior measures agreement with the specified bounds before the update. The construction turns admissibility from a post-processing step into conditioning information for probabilistic PDE solvers.