Self-attention summary networks for subsurface velocity-model building from common-image gathers

Key points

• Common-image gathers (CIGs) contain physically meaningful information about velocity-model errors through reflector focusing and residual moveout, but in conventional imaging workflows they are typically used only as diagnostic tools.1
• These learned embeddings preserve offset-dependent kinematic structure and spatial coherence while reducing variability caused by background-velocity mismatch.1
• Conditioned on these summary embeddings, a flow-matching model learns a transport from a Gaussian source distribution to the posterior distribution of plausible velocity fields.1
• Numerical experiments show that, compared with direct conditioning on raw CIGs, the proposed summary network improves posterior velocity inference.1

Sources (1)

• [1]Self-attention summary networks for subsurface velocity-model building from common-image gathers
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 01:31 AM

  In this work, we propose a multiscale self-attention summary network that maps high-dimensional 3D CIG volumes into compact conditioning embeddings for probabilistic subsurface velocity inversion.

  Common-image gathers (CIGs) contain physically meaningful information about velocity-model errors through reflector focusing and residual moveout, but in conventional imaging workflows they are typically used only as diagnostic tools.

Extractive summary: sentences quoted from the sources.