Research map / Probabilistic and scientific modeling / Variational and generative sampling
DeepRWCap: Neural-Guided Random-Walk Capacitance Solver for IC Design
Hector Rodriguez Rodriguez, Jiechen Huang, Wenjian Yu
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 grid positional encoding to the depthwise separable face predictor (ablation) | DeepRWCap Poisson face predictor · Poisson kernel prediction | L2 error (%): 12.15 -> 3.93 | Improved |
| use fixed grid PE instead of learned PE (ablation) | DS Conv face predictor · Poisson kernel prediction | L2 error (%): 4.11 -> 3.93 | Improved |
| use depthwise separable conv instead of standard 2D conv (ablation) | face predictor · Poisson kernel prediction | L2 error (%): 13.80 -> 12.15 | Improved |
| decompose into face selection plus 2D face prediction instead of 3D conv (ablation) | face predictor · Poisson kernel prediction | L2 error (%): 24.15 -> 12.15 | Improved |
| replace MLP with DS Conv (ablation) | face predictor · Poisson kernel prediction | L2 error (%): 7.83 -> 12.15 (no PE); 3.93 with grid PE; params 34.4 M -> 1.37 K | Improved |
| neural-guided transitions (DeepRWCap) instead of GE-CNN guidance | FRW capacitance extraction · Cases 9-10 (12 nm) | relative error (%): 22.9 -> 1.1 (case 9); 27.1 -> 1.2 (case 10) | Improved |
| DeepRWCap versus Microwalk | FRW capacitance extraction · 10 industrial designs | runtime: 1.23x speedup (p = 0.024) | Improved |
| DeepRWCap versus FRW-AGF | FRW capacitance extraction · 10 industrial designs | mean relative error (%): 5.18% ± 7.81% -> 1.24% ± 0.53% | Mixed |
| TensorRT and custom CUDA kernels (ablation) | DeepRWCap inference · batch 2048, Poisson / gradient | execution time (ms): 15.2 -> 2.9 ms; 32.1 -> 3.8 ms | Improved |
Closest papers by meaning
- DeepGate4: Efficient and Effective Representation Learning for Circuit Design at Scale (ICLR 2025 · similarity 0.83)
- NeuralSteiner: Learning Steiner Tree for Overflow-avoiding Global Routing in Chip Design (NeurIPS 2024 · similarity 0.83)
- DeepLayout: Learning Neural Representations of Circuit Placement Layout (ICML 2025 · similarity 0.83)
- CircuitFusion: Multimodal Circuit Representation Learning for Agile Chip Design (ICLR 2025 · similarity 0.82)
- FlexPlanner: Flexible 3D Floorplanning via Deep Reinforcement Learning in Hybrid Action Space with Multi-Modality Representation (NeurIPS 2024 · similarity 0.82)
- Co-PiLOT: Constrained Physics-Informed Latent Optimization for Target-Driven Inverse Design (NeurIPS 2026 · similarity 0.81)
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.
