Including Physics in Deep Learning -- An example from 4D seismic pressure saturation inversion
This work addresses data uncertainty and anomaly detection in geoscience, offering a domain-specific solution for improved seismic inversion.
The authors tackled the problem of imbalanced learning in geoscience data by incorporating physical priors into neural network architectures and using noise injection to transfer models from synthetic to field data, achieving successful application in 4D seismic pressure saturation inversion.
Geoscience data often have to rely on strong priors in the face of uncertainty. Additionally, we often try to detect or model anomalous sparse data that can appear as an outlier in machine learning models. These are classic examples of imbalanced learning. Approaching these problems can benefit from including prior information from physics models or transforming data to a beneficial domain. We show an example of including physical information in the architecture of a neural network as prior information. We go on to present noise injection at training time to successfully transfer the network from synthetic data to field data.