Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Proximal Splitting Methods for Hybrid Differentiable Models
Abdel-Rahim Mezidi, Jordan Patracone, Amaury Habrard
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 |
|---|---|---|---|
| use HybDR instead of SITL+ in the solver-correction pipeline | Allen-Cahn Diffusion Only · Allen-Cahn (Diffusion Only, ResNet) | test MSE: 9.03 × 10−3 -> 5.70 × 10−3 | Improved |
| use HybDR instead of SITL | Allen-Cahn Diffusion Only · Allen-Cahn (Diffusion Only, ResNet) | test MSE: 2.62 × 10−1 -> 5.70 × 10−3 | Improved |
| add learned neural correction inside the solver loop (HybDR) over solver-only | Allen-Cahn Both Terms · Allen-Cahn (Both Terms, ResNet) | test MSE: 2.06 × 10−1 -> 4.72 × 10−3 | Improved |
| use HybDR in Reaction Only setting where SITL and SITL+ fail | Allen-Cahn Reaction Only · Allen-Cahn (Reaction Only, ResNet) | convergence / test MSE: HybDR stable at 5.42 × 10−3 while SITL and SITL+ failed to converge | Improved |
| replace ResNet with FNO as the network (ablation) | HybDR / HybFB-post on Allen-Cahn · Allen-Cahn (FNO) | test MSE | No clear effect |
| replace ResNet with FNO as the network (ablation) | HybDR / HybFB-post on Kuramoto-Sivashinsky · Kuramoto-Sivashinsky (FNO) | test MSE | Mixed |
| increase spatial resolution from 128 to 2048 (ablation) | HybFB-post and HybDR on Kuramoto-Sivashinsky · Kuramoto-Sivashinsky resolution sweep | test MSE: only a mild increase at resolution 2048 | No clear effect |
Closest papers by meaning
- Holistic Physics Solver: Learning PDEs in a Unified Spectral-Physical Space (ICML 2025 · similarity 0.85)
- HEAP: Hyper Extended A-PDHG Operator for Constrained High-dim PDEs (ICML 2025 · similarity 0.84)
- Using Uncertainty Quantification to Characterize and Improve Out-of-Domain Learning for PDEs (ICML 2024 · similarity 0.84)
- Mosaic: A Benchmark Suite for Differentiable Physics Solvers (NeurIPS 2026 · similarity 0.84)
- Accelerating PDE-Constrained Optimization by the Derivative of Neural Operators (ICML 2025 · similarity 0.84)
- Data-Efficient Operator Learning via Unsupervised Pretraining and In-Context Learning (NeurIPS 2024 · 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.
