Research map / Probabilistic and scientific modeling / Variational and generative sampling
DeepGEM: Generalized Expectation-Maximization for Blind Inversion
Angela F Gao, Jorge Castillo Castellanos, Yisong Yue, Zachary E Ross, Katherine Bouman
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase number of sources from 9 to 100 (ablation) | DeepGEM · Blind seismic tomography, random source configs | velocity error (km/s): 0.52 ± 0.063 -> 0.27 ± 0.021 | Improved |
| increase number of sources from 9 to 100 (ablation) | DeepGEM · Blind seismic tomography, random source configs | source localization error (m): 62.88 ± 21.77 -> 30.38 ± 1.43 | Improved |
| increase number of sources from 25 to 49 (ablation) | DeepGEM · Blind seismic tomography, random source configs | velocity error (km/s): 0.42 ± 0.043 -> 0.44 ± 0.181 | No clear effect |
| add homogeneous Lθ prior instead of none (ablation) | DeepGEM · Blind seismic tomography, random velocity fields | velocity error (km/s): 0.51 ± 0.30 -> 0.94 ± 0.45 | Hurt |
| add gradient Lθ prior instead of none (ablation) | DeepGEM · Blind seismic tomography, random velocity fields | velocity error (km/s): 0.51 ± 0.30 -> 1.07 ± 0.53 | Hurt |
| add homogeneous Lθ prior instead of none (ablation) | DeepGEM · Blind seismic tomography, random velocity fields | source localization error (m): 24.40 ± 7.18 -> 21.37 ± 5.34 | Improved |
| add gradient Lθ prior instead of none (ablation) | DeepGEM · Blind seismic tomography, random velocity fields | source localization error (m): 24.40 ± 7.18 -> 26.20 ± 6.46 | Hurt |
| use perfectly known source locations instead of uncertain ones (ablation) | DeepGEM · Blind seismic tomography, random velocity fields | velocity error (km/s): 0.51 ± 0.30 -> 0.57 ± 0.25 | No clear effect |
| use gradient or layer Lθ prior closer to true velocity (ablation) | DeepGEM · Blind seismic tomography, single velocity model | velocity MSE | Improved |
| single receiver with perfectly known source locations and noisy measurements (ablation) | DeepGEM · Blind seismic tomography, one receiver | reconstruction artifacts | Hurt |
| use noise-free travel time measurements with a single receiver (ablation) | DeepGEM · Blind seismic tomography, one receiver | reconstruction artifacts | Improved |
| DeepGEM MAPθ instead of MAPθ,x | DeepGEM · Blind seismic tomography | velocity MSE | Improved |
Closest papers by meaning
- A Unified Framework for Forward and Inverse Problems in Subsurface Imaging using Latent Space Translations (ICLR 2025 · similarity 0.87)
- An Intriguing Property of Geophysics Inversion (ICML 2022 · similarity 0.86)
- SGD Jittering: A Training Strategy for Robust and Accurate Model-Based Architectures (ICML 2025 · similarity 0.85)
- EasyInv: Toward Fast and Better DDIM Inversion (ICML 2025 · similarity 0.84)
- Integrating Intermediate Layer Optimization and Projected Gradient Descent for Solving Inverse Problems with Diffusion Models (ICML 2025 · similarity 0.84)
- NPN: Non-Linear Projections of the Null-Space for Imaging Inverse Problems (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
