DINGO for lensed signals

Code purpose

Machine-learning inference · in development

Simulation-based inference built on DINGO makes large-scale lensing analyses tractable. A model currently exists for the isolated point lens.

Contributors

Srashti Goyal (lead developer).

Lensing searches face a growing computational problem. Every wave-optics parameter added to a waveform enlarges the space an analysis has to explore, and the detector network will soon deliver thousands of events a year^[1] , with the cost rising again for searches over pairs of events, where multiply-imaged candidates have to be compared with one another. A network trained to map data directly to posteriors amortizes that cost: the expense moves into training, and each subsequent event is analyzed in seconds.^[2]

The models are built on DINGO, a neural-posterior-estimation framework for gravitational-wave inference^[3] that is developed and maintained outside this project. What the project adds is the lensing. A model exists today for the isolated point lens, trained on amplification factors computed with GLoW, and the two pieces are meant to fit together: the wave-optics code supplies the training data, the network supplies the speed. Why this matters for the searches themselves is argued under data analysis.

Availability

There is no official release of the lensing models. DINGO itself is public and maintained outside the project, at github.com/dingo-gw/dingo, and what the project adds to it lives with the papers below for now. The release plans say what the intention is.

Citing it

Work using these models should cite DINGO itself, and the paper that brought the lensing in and demonstrated it on an isolated point lens.^[4]

• Real-time gravitational wave science with neural posterior estimation M. Dax, S. R. Green, J. Gair et al. · Phys. Rev. Lett. 127, 241103 (2021) · arXiv:2106.12594
• Accelerated inference of microlensed gravitational waves with machine learning M. Caldarola, S. Goyal, N. Gupte et al. · Phys. Rev. D 113, 104073 (2026) · arXiv:2511.08486

References

1. LIGO Scientific Collaboration, LIGO–Virgo–KAGRA observing plan Observing capabilities documentation
2. K. Cranmer, J. Brehmer, G. Louppe, The frontier of simulation-based inference (2019)
3. M. Dax, S. R. Green, J. Gair et al., Real-Time Gravitational Wave Science with Neural Posterior Estimation Phys. Rev. Lett. 127, 241103 (2021)
4. M. Caldarola, S. Goyal, N. Gupte, S. R. Green, M. Zumalacárregui, Accelerated inference of microlensed gravitational waves with machine learning Phys. Rev. D 113, 104073 (2026)

Other packages