The Double Emulator
Conor Crilly, Oliver Johnson, Alexander Lewis, Jonathan Rougier

TL;DR
This paper introduces the double emulator, a modified Gaussian process emulator designed to better model simulators with non-stationary and non-smooth outputs, demonstrated on corrosion simulation data.
Contribution
The paper proposes the double emulator, an adaptation of GPEs, to handle simulators that ground in large input regions, addressing limitations of traditional GPEs.
Findings
Double emulator outperforms GPE on non-stationary simulators
Improved accuracy in modeling corrosion simulation
Effective on synthetic and real-world examples
Abstract
Computer models (simulators) are vital tools for investigating physical processes. Despite their utility, the prohibitive run-time of simulators hinders their direct application for uncertainty quantification. Gaussian process emulators (GPEs) have been used extensively to circumvent the cost of the simulator and are known to perform well on simulators with smooth, stationary output. In reality, many simulators violate these assumptions. Motivated by a finite element simulator which models early stage corrosion of uranium in water vapor, we propose an adaption of the GPE, called the double emulator, specifically for simulators which 'ground' in a considerable volume of their input space. Grounding is the process by which a simulator attains its minimum and can result in violation of the stationarity and smoothness assumptions used in the conventional GPE. We perform numerical…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsProbabilistic and Robust Engineering Design · Model Reduction and Neural Networks · Advanced Multi-Objective Optimization Algorithms
