Gravitational-wave parameter estimation with autoregressive neural network flows
read the original abstract
We introduce the use of autoregressive normalizing flows for rapid likelihood-free inference of binary black hole system parameters from gravitational-wave data with deep neural networks. A normalizing flow is an invertible mapping on a sample space that can be used to induce a transformation from a simple probability distribution to a more complex one: if the simple distribution can be rapidly sampled and its density evaluated, then so can the complex distribution. Our first application to gravitational waves uses an autoregressive flow, conditioned on detector strain data, to map a multivariate standard normal distribution into the posterior distribution over system parameters. We train the model on artificial strain data consisting of IMRPhenomPv2 waveforms drawn from a five-parameter $(m_1, m_2, \phi_0, t_c, d_L)$ prior and stationary Gaussian noise realizations with a fixed power spectral density. This gives performance comparable to current best deep-learning approaches to gravitational-wave parameter estimation. We then build a more powerful latent variable model by incorporating autoregressive flows within the variational autoencoder framework. This model has performance comparable to Markov chain Monte Carlo and, in particular, successfully models the multimodal $\phi_0$ posterior. Finally, we train the autoregressive latent variable model on an expanded parameter space, including also aligned spins $(\chi_{1z}, \chi_{2z})$ and binary inclination $\theta_{JN}$, and show that all parameters and degeneracies are well-recovered. In all cases, sampling is extremely fast, requiring less than two seconds to draw $10^4$ posterior samples.
This paper has not been read by Pith yet.
Forward citations
Cited by 4 Pith papers
-
Parameter inference of millilensed gravitational waves using neural spline flows
Neural spline flows perform fast posterior inference on 11-dimensional millilensed GW parameters with accuracy comparable to dynesty for most quantities and a 3-day to 0.8-second speedup.
-
Artifact-Conditioned Interval Diagnostics for Flow-Matching Neural Posterior Estimation in a Controlled Gravitational-Wave Benchmark
In a controlled binary-black-hole benchmark, soft learned artifact-aware interval rescaling (LAIR) reduces marginal calibration error for frequency masks from 0.1195 to 0.0672 but is not uniformly better than raw inte...
-
Assessment of normalizing flows for parameter estimation on time-frequency representations of gravitational-wave data
GP15 maps BBH spectrograms to parameter posteriors via residual networks and normalizing flows, producing results consistent with LVK analyses on GWTC-2.1 and GWTC-3 events while running in seconds.
-
Not too close! Evaluating the impact of the baseline on the localization of binary black holes by next-generation gravitational-wave detectors
Baselines of 8-11 ms light travel time for two CE detectors provide a reasonable compromise for BBH sky localization, with third detectors eliminating multimodality for most or all events.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.