Accelerating physics-informed neural networks for full waveform inversion using a hybrid quantum-classical finite-basis architecture
Hoang Anh Nguyen, Divakar Vashisth, and Ali Tura · 2026
Read Accelerating physics-informed neural networks for full waveform inversion using a hybrid quantum-classical finite-basis architecture by Hoang Anh Nguyen, Divakar Vashisth, and Ali Tura (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-01110.txt.
Description
Nguyen, Vashisth, and Tura put quantum circuits inside a seismic imaging network in this 2026 paper. Full waveform inversion rebuilds underground properties from receiver data. It is powerful and costly. Physics informed neural networks offer a mesh free rival, but they strain to fit complex velocity fields. Domain split variants help, at the price of harder convergence.
Their hybrid makes both halves quantum aware. The wavefield network and the global velocity network end in parameterized quantum circuits. The circuits run as differentiable simulators in JAX, so gradients flow end to end through physics loss, circuit, and classical nets alike. On an anomaly benchmark, the hybrid reaches a lower velocity error than the main classical baseline. It does so in about eight times fewer training steps, with a third fewer parameters. It also beats all 15 classical variants tested. A checkerboard test shows the pipeline handles broad patterns, not just one anomaly.
Catalog record
| CREATOR | Hoang Anh Nguyen, Divakar Vashisth, and Ali Tura |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.01110