Integrate GPry framework to PyCBC for accelerated Bayesian inference

[jahedabedi]

Standard information about the request

This is a: new feature

This change affects: inference

This change changes: scientific output

This change: follows style guidelines, requires additional dependencies

This change will: require additional dependencies (GPry)

Motivation

With the increasing rate of detected gravitational wave events (90+ in O3) and next-generation detectors like LISA and Einstein Telescope, traditional MCMC methods —reliant on iterative waveform evaluations with non-negligible computational cost per likelihood calculation—face fundamental scalability limitations. GPry accelerates Bayesian inference using Gaussian Process Regression and active learning, achieving:

• 100x acceleration factor (𝒜 = t_traditional/t_GPry)
• Significant reduction in CO₂ emissions per analysis
• Speeding up gating-based parameter estimation and making ringdown analyses far more tractable.

This enables efficient parameter estimation and prepares PyCBC for cosmic explorer-era data rates.

Contents

• Core Integration:

  • New GPrySampler class in pycbc.inference.sampler
  • Active learning interface with PyCBC's waveform generators

• Documentation:

  • GPry source code on GitHub
  • GPry Documentation on Read the Docs
  • arXiv:2211.02045 for the core algorithm.

• The author confirms adherence to the code of conduct

[cdcapano]

Adding the WIP label as Jahed is working to update this to the current version of GPry.