Autoencoder model for fast generation of effective one-body gravitational waveform approximations
快速生成有效单体引力波波形近似的自编码器模型
Upgrades to current gravitational wave detectors for the next observation run and upcoming third-generation observatories, like the Einstein telescope, are expected to have enormous improvements in detection sensitivities and compact object merger event rates. Estimation of source parameters for a wider parameter space that these detectable signals will lie in will be a computational challenge. Thus, it is imperative to have methods to speed up the likelihood calculations with theoretical waveform predictions, which can ultimately make the parameter estimation faster and aid in rapid multimessenger follow-ups. In this work, we study autoencoder models for gravitational waveform generation by adopting the best-performing architecture of Liao and Lin [1] to approximate fixed duration 1 s long aligned-spin SEOBNRv4 inspiral-merger-ringdown waveforms. Our parameter space consists of four parameters, [<inline-formula><mml:math><mml:mrow><mml:msub><mml:mrow><mml:mi>m</mml:mi></mml:mrow><mml:mrow><mml:mn>1</mml:mn></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math><mml:mrow><mml:msub><mml:mrow><mml:mi>m</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math><mml:msub><mml:mi>χ</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>(</mml:mo><mml:mi>z</mml:mi><mml:mo>)</mml:mo></mml:math></inline-formula>, <inline-formula><mml:math><mml:mrow><mml:msub><mml:mrow><mml:mi>χ</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msub><mml:mo>(</mml:mo><mml:mi>z</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula>]. The masses are uniformly sampled in <inline-formula><mml:math><mml:mrow><mml:mo>[</mml:mo><mml:mn>5</mml:mn><mml:mo>,</mml:mo><mml:mn>75</mml:mn><mml:mo>]</mml:mo><mml:msub><mml:mi>M</mml:mi><mml:mo>⊙</mml:mo></mml:msub></mml:mrow></mml:math></inline-formula> with a mass ratio limit of <inline-formula><mml:math><mml:msub><mml:mi>m</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>/</mml:mo><mml:msub><mml:mi>m</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo><</mml:mo><mml:mn>10</mml:mn></mml:math></inline-formula>, while the spins are uniform in <inline-formula><mml:math><mml:mo>[</mml:mo><mml:mo>-</mml:mo><mml:mn>0.99</mml:mn><mml:mo>,</mml:mo><mml:mn>0.99</mml:mn><mml:mo>]</mml:mo></mml:math></inline-formula>. Our model is able to generate <inline-formula><mml:math><mml:msup><mml:mn>10</mml:mn><mml:mn>3</mml:mn></mml:msup></mml:math></inline-formula> waveforms in about <inline-formula><mml:math><mml:mrow><mml:msup><mml:mrow><mml:mn>10</mml:mn></mml:mrow><mml:mrow><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:mtext> </mml:mtext><mml:mtext> </mml:mtext><mml:mi>s</mml:mi></mml:mrow></mml:math></inline-formula> at an average speed of <inline-formula><mml:math><mml:mrow><mml:mn>50</mml:mn><mml:mtext> </mml:mtext><mml:mtext> </mml:mtext><mml:mi>μs</mml:mi></mml:mrow></mml:math></inline-formula> per waveform on a graphics processing unit (GPU). This batched GPU generation is about 4 orders of magnitude faster than the serialized CPU generation by the base SEOBNRv4 implementation, and 2─3 orders of magnitude faster than existing non-machine-learning accelerated waveform variants. The median mismatch for the generated waveforms in the test dataset is <inline-formula><mml:math><mml:mrow><mml:mo>∼</mml:mo><mml:msup><mml:mn>10</mml:mn><mml:mrow><mml:mo>-</mml:mo><mml:mn>2</mml:mn></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula>, with better accuracy in a restricted parameter space of <inline-formula><mml:math><mml:msub><mml:mi>χ</mml:mi><mml:mi>eff</mml:mi></mml:msub><mml:mo>∈</mml:mo><mml:mo>[</mml:mo><mml:mo>-</mml:mo><mml:mn>0.80</mml:mn><mml:mo>,</mml:mo><mml:mn>0.80</mml:mn><mml:mo>]</mml:mo></mml:math></inline-formula>, however still less than that of other approaches. The latent sampling error of our model can be quantified at a median mismatch standard deviation of <inline-formula><mml:math><mml:mn>4</mml:mn><mml:mo>×</mml:mo><mml:msup><mml:mn>10</mml:mn><mml:mrow><mml:mo>-</mml:mo><mml:mn>3</mml:mn></mml:mrow></mml:msup></mml:math></inline-formula>. Although the accuracy of our model does not enable full production use yet, the model could be useful wherever a high volume of approximate theoretical waveforms are required, for instance, for rapid sky localization.
展开 ▾首次将对齐自旋的完整inspiral-merger-ringdown波形纳入自编码器生成框架,在GPU上实现约50微秒/波形的超高生成速度,远超现有加速波形方案,揭示了机器学习在引力波波形快速近似中的潜力。
随着引力波探测器灵敏度提升和事件数量激增,快速理论波形生成成为参数估计瓶颈。尽管机器学习在该领域已有诸多尝试,但大多局限于小参数空间或部分波形阶段(如 Cuoco+ 2025)。Liao 和 Lin 首次用条件自编码器生成零自旋 inspiral-merger 波形 Liao+ 2021,本研究在此基础上扩展至对齐自旋的完整 IMR 场景,并引入动态低截断频率以标准化输入长度。模型生成速度比原生 SEOBNRv4 快约四个数量级,较优化频域 ROM 波形 Bohé+ 2016 也快 2–3 个数量级,但精度尚不足以替代标准波形(中位失配约10^{-2})。未来可通过改进网络设计、融入失配感知损失、扩展参数空间(进动、偏心率)及结合快速似然方法,逐步迈向生产级的实时引力波参数估计。
预印本 2025-11-16 · 刊出 2026-07-16 · 收录 2026-07-27