Research map / Probabilistic and scientific modeling / Differential equations and PDEs
HDK: Toward High-Performance Deep-Learning-Based Kirchhoff Analysis
Xinying Wang, Olamide Timothy Tawose, Feng Yan, Dongfang Zhao
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 |
|---|---|---|---|
| parallelize forward labelling with MPI (64 threads) versus serial | HDK labelling · 64x64 array, 126 equations | labelling speedup: 6,787 seconds -> 8.7× | Improved |
| increase threads beyond 16 for parallel labelling | HDK labelling · 64x64 array | speedup | No clear effect |
| add PCA dimension reduction to 1,024 dimensions (ablation) | HDK · cell medium | accuracy: 96.7% -> around 99% | Improved |
| add PCA dimension reduction to 1,024 dimensions (ablation) | HDK · wound surface | accuracy: 94.7% -> around 99% | Improved |
| reduce dimensions too far (15 dims) (ablation) | HDK · cell medium | accuracy: 99.3% -> 96.8% | Hurt |
| reduce dimensions too far (19 dims) versus 32 and 64 (ablation) | HDK · wound surface | accuracy: 99.4% -> 97.8% | Hurt |
| increase PCA dimensions from 32 to 64 (ablation) | HDK · wound surface | accuracy: 98.9% -> 99.4% | Improved |
| augment training data from 100 to 800 points with random errors (ablation) | HDK · cell medium | accuracy: 87.8% -> 99% | Improved |
| augment training data from 100 to 800 points with random errors (ablation) | HDK · wound surface | accuracy: 91.8% -> 99% | Improved |
| augment beyond 800 data points (ablation) | HDK · both data sets | accuracy | Hurt |
| HDK versus CNN-based Kirchhoff analysis (CKA) on accuracy | HDK vs CKA · 64x64 array | accuracy: 92.5% -> 99.6% | Improved |
| HDK versus CNN-based Kirchhoff analysis (CKA) on time | HDK vs CKA · 64x64 array | end-to-end time: 2.1 hours -> 15 minutes | Improved |
Closest papers by meaning
- A Deep Conjugate Direction Method for Iteratively Solving Linear Systems (ICML 2023 · similarity 0.82)
- Poisson-Dirac Neural Networks for Modeling Coupled Dynamical Systems across Domains (ICLR 2025 · similarity 0.80)
- Physics-Informed DeepONets for drift-diffusion on metric graphs: simulation and parameter identification (ICML 2025 · similarity 0.80)
- Schur's Positive-Definite Network: Deep Learning in the SPD cone with structure (ICLR 2025 · similarity 0.80)
- Investigating the Overlooked Hessian Structure: From CNNs to LLMs (ICML 2025 · similarity 0.80)
- Co-PiLOT: Constrained Physics-Informed Latent Optimization for Target-Driven Inverse Design (NeurIPS 2026 · 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.
