Research map / Probabilistic and scientific modeling / Dynamical systems
BHCast: Unlocking Black Hole Plasma Dynamics from a Single Blurry Image with Long-Term Forecasting
Renbo Tu, Ali SaraerToosi, Nicholas S. Conroy, Gennady Pekhimenko, Aviad Levis
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| Add multi-scale Laplacian pyramid loss with mean-flux term instead of L2 only (ablation) | BHCast U-Net forecaster · Sgr A* forecasting LPIPS | LPIPS stability over rollout | Improved |
| Add mean-flux regularization to the multi-scale pyramid loss (ablation) | BHCast U-Net forecaster · Sgr A* forecasting LPIPS | LPIPS stability over rollout: stable only for first 20 forecasts without mean flux | Improved |
| Forecast dynamics then extract plasma features instead of direct ResNet regression | BHCast · Sgr A* blurred input | Pattern speed MAE: 0.64 -> 0.46 | Improved |
| Forecast then extract features instead of direct ResNet regression | BHCast · Sgr A* blurred input | Asymmetry MAE: 0.23 -> 0.30 | Hurt |
| Use forecast-based features with XGBoost instead of direct ResNet classification | BHCast · Sgr A* inclination, 20 µas | accuracy: 47.19% -> 56.41% | Improved |
| Increase input blur to 30 µas | ResNet baseline · Sgr A* inclination | accuracy: 47.19% -> 25.47% | Hurt |
| Increase input blur to 30 µas | BHCast · Sgr A* inclination | accuracy: 56.41% -> 53.44% | No clear effect |
| Increase input blur to 25 µas | BHCast · Sgr A* spin | accuracy: 69.22% -> 71.09% | No clear effect |
| Autoregressive forecasting with U-Net replaces decoupled EDSR + ConvLSTM baseline | BHCast · Sgr A* 100-step forecast | LPIPS | Improved |
Closest papers by meaning
- Probabilistic Data-Driven Modelling of Astrophysical Transients: The Neural Process Family for Ultrafast and Class-Agnostic Light Curve Reconstruction (NeurIPS 2026 · similarity 0.82)
- Space and time continuous physics simulation from partial observations (ICLR 2024 · similarity 0.80)
- BERTground: A Transformer-Based Model of Background Spectra on the ISS-Based NICER Space Telescope (AAAI 2024 · similarity 0.80)
- Generating Full-field Evolution of Physical Dynamics from Irregular Sparse Observations (NeurIPS 2025 · similarity 0.79)
- VESTA: Visual Exploration with Statistical Tool Agents (NeurIPS 2026 · similarity 0.79)
- Dynamic Diffusion Schrödinger Bridge in Astrophysical Observational Inversions (NeurIPS 2025 · similarity 0.79)
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.
