Abstract:Uncertainty quantification (UQ) in scientific machine learning (SciML) becomes increasingly critical as neural networks (NNs) are being widely adopted in addressing complex problems across various scientific disciplines. Representative SciML models are physics-informed neural networks (PINNs) and neural operators (NOs). While UQ in SciML has been increasingly investigated in recent years, very few works have focused on addressing the uncertainty caused by the noisy inputs, such as spatial–temporal coordinates in PINNs and input functions in NOs. The presence of noise in the inputs of the models can pose significantly more challenges compared to noise in the outputs of the models, primarily due to the inherent nonlinearity of most SciML algorithms. As a result, UQ for noisy inputs becomes a crucial factor for reliable and trustworthy deployment of these models in applications involving physical knowledge. To this end, we introduce a Bayesian approach to quantify uncertainty arising from noisy inputs–outputs in PINNs and NOs. We show that this approach can be seamlessly integrated into PINNs and NOs, when they are employed to encode the physical information. PINNs incorporate physics by including physics-informed terms via automatic differentiation, either in the loss function or the likelihood, and often take as input the spatial–temporal coordinate. Therefore, the present method equips PINNs with the capability to address problems where the observed coordinate is subject to noise. On the other hand, pretrained NOs are also commonly employed as equation-free surrogates in solving differential equations and Bayesian inverse problems, in which they take functions as inputs. The proposed approach enables them to handle noisy measurements for both input and output functions with UQ. We present a series of numerical examples to demonstrate the consequences of ignoring the noise in the inputs and the effectiveness of our approach in addressing noisy inputs–outputs with UQ when PINNs and pretrained NOs are employed for physics-informed learning.

Uncertainty-quantification-enabled inversion of nuclear responses

Uncertainty Quantification-Enabled Inversion of Nuclear Euclidean Responses

Invertible neural networks for uncertainty quantification in refraction tomography

Uncertainty Quantification for Nuclear Density Functional Theory and Information Content of New Measurements

Uncertainty quantification for noisy inputs-outputs in physics-informed neural networks and neural operators

Deep Neural Network Uncertainty Quantification for LArTPC Reconstruction

Parameterizing uncertainty by deep invertible networks, an application to reservoir characterization

Universal uncertainty estimation for nuclear detector signals with neural networks and ensemble learning

Quantifying total uncertainty in physics-informed neural networks for solving forward and inverse stochastic problems

Uncertainty quantification in the machine-learning inference from neutron star probability distribution to the equation of state

Uncertainty quantification for noisy inputs–outputs in physics-informed neural networks and neural operators

Uncertainty Estimation for Molecules: Desiderata and Methods

Uncertainty Quantification in Seismic Inversion Through Integrated Importance Sampling and Ensemble Methods

Functional PCA and deep neural networks-based Bayesian inverse uncertainty quantification with transient experimental data

Neural State-Space Models: Empirical Evaluation of Uncertainty Quantification

Machine learning the deuteron: new architectures and uncertainty quantification

A Kernel Framework to Quantify a Model's Local Predictive Uncertainty under Data Distributional Shifts

UncertaINR: Uncertainty Quantification of End-to-End Implicit Neural Representations for Computed Tomography

Quantifying Model Uncertainty of Neural Network-based Turbulence Closures