Dingo

Dingo (Deep Inference for Gravitational-wave Observations) is a Python program for analyzing gravitational wave data using neural posterior estimation. It dramatically speeds up inference of astrophysical source parameters from data measured at gravitational-wave observatories. Dingo aims to enable the routine use of the most advanced theoretical models in analysing data, to make rapid predictions for multi-messenger counterparts, and to do so in the context of sensitive detectors with high event rates.

The basic approach of Dingo is to train a neural network to represent the Bayesian posterior, conditioned on data. This enables amortized inference: when new data are observed, they can be plugged in and results obtained in a small amount of time. Tasks handled by Dingo include

• building training datasets;
• training normalizing flows to estimate the posterior density;
• performing inference on real or simulated data; and
• verifying and correcting model results using importance sampling.

As training a network from scratch can be expensive, we intend to also distribute trained networks that can be used directly for inference. These can be used with dingo_pipe to automate analysis of gravitational wave events.

References

Dingo is based on a series of papers developing neural posterior estimation for gravitational waves, starting from proof of concept (Green et al. 2020), to inclusion of all 15 parameters and analysis of real data (Green and Gair 2021), noise conditioning and full amortization (Dax et al. 2021), and group-equivariant NPE (Dax, Green, Gair, Deistler, et al. 2022). Dingo results are augmented with importance sampling in (Dax, Green, Gair, Pürrer, et al. 2022). Finally, training with forecasted noise (needed for training prior to an observing run) is described in (Wildberger et al. 2022).

Dax, Maximilian, Stephen R. Green, Jonathan Gair, Michael Pürrer, et al. 2022. Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inference. October. https://arxiv.org/abs/2210.05686.

Dax, Maximilian, Stephen R. Green, Jonathan Gair, Michael Deistler, Bernhard Schölkopf, and Jakob H. Macke. 2022. “Group equivariant neural posterior estimation.” International Conference on Learning Representations. https://arxiv.org/abs/2111.13139.

Dax, Maximilian, Stephen R. Green, Jonathan Gair, Jakob H. Macke, Alessandra Buonanno, and Bernhard Schölkopf. 2021. “Real-Time Gravitational Wave Science with Neural Posterior Estimation.” Phys. Rev. Lett. 127 (24): 241103. https://doi.org/10.1103/PhysRevLett.127.241103.

Green, Stephen R., and Jonathan Gair. 2021. “Complete parameter inference for GW150914 using deep learning.” Mach. Learn. Sci. Tech. 2 (3): 03LT01. https://doi.org/10.1088/2632-2153/abfaed.

Green, Stephen R., Christine Simpson, and Jonathan Gair. 2020. “Gravitational-wave parameter estimation with autoregressive neural network flows.” Phys. Rev. D 102: 104057. https://doi.org/10.1103/PhysRevD.102.104057.

Wildberger, Jonas, Maximilian Dax, Stephen R. Green, et al. 2022. Adapting to noise distribution shifts in flow-based gravitational-wave inference. November. https://arxiv.org/abs/2211.08801.

If you use Dingo in your work, we ask that you please cite at least (Dax et al. 2021).

Contributors to the code are listed in AUTHORS.md. We thank Vivien Raymond and Rory Smith for acting as LIGO-Virgo-KAGRA (LVK) code reviewers. Dingo makes use of many LVK software tools, including Bilby, bilby_pipe, and LALSimulation, as well as third party tools such as PyTorch and nflows.

Contact

For questions or comments please contact Maximilian Dax or Stephen Green.