Research map / Learning theory and algorithms / Kernels and regression theory
Characterizing and Optimizing the Spatial Kernel of Multi Resolution Hash Encodings
Tianxiang Dai, Jonathan Fan
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 |
|---|---|---|---|
| Apply a distinct rotation per resolution level (R-MHE, M=8) | MHE · 2D image regression avg | PSNR (dB): 23.88 -> 24.82 | Improved |
| Increase effective rotations from M=2 to M=8 (ablation) | R-MHE · 2D image regression avg | PSNR (dB): 24.62 -> 24.82 | Improved |
| Use R-MHE Icosa rotations with baseline-heuristic b | MHE · Synthetic NeRF (8 scenes) | PSNR (dB): 35.346 -> 35.479 | No clear effect |
| Use icosahedron instead of tetrahedron rotations with PSF-guided b (ablation) | R-MHE · Synthetic NeRF (8 scenes) | PSNR (dB): 35.396 -> 35.440 | Improved |
| Use PSF-guided theory b instead of empirically optimized b (ablation) | Standard MHE · Synthetic NeRF (8 scenes) | PSNR (dB): 35.346 -> 35.329 | No clear effect |
| Add R-MHE rotations | MHE · SDF (Armadillo, Bunny, Spot) | IoU: 0.9986 -> 0.9986 | No clear effect |
| Lower growth factor b from b_theory to b_theory-0.1 (ablation) | Standard MHE · Mountain image regression | PSNR (dB): 23.07 -> 24.19 | Improved |
| Raise growth factor b from b_theory to b_theory+0.2 (ablation) | Standard MHE · Mountain image regression | PSNR (dB): 23.07 -> 21.76 | Hurt |
| Lower b from b_theory-0.1 to b_theory-0.2 (ablation) | R-MHE M=8 · City image regression | PSNR (dB): 22.43 -> 22.82 | Improved |
| Increase M from 1 to larger values (ablation) | R-MHE · PSF isotropy | anisotropy ratio | Mixed |
| Increase number of levels L or growth factor b (ablation) | MHE · hash collision SNR at fixed T | SNR | Improved |
| Optimize with Adam instead of the minimum-norm ideal solution | MHE PSF · PSF broadening | broadening factor: 1.18 -> 3.0 | Hurt |
Closest papers by meaning
- Grid4D: 4D Decomposed Hash Encoding for High-Fidelity Dynamic Gaussian Splatting (NeurIPS 2024 · similarity 0.82)
- HashPoint: Accelerated Point Searching and Sampling for Neural Rendering (CVPR 2024 · similarity 0.82)
- TwinPrune: Density-Aware Two-Phase Token Pruning for Vision-Language Models (NeurIPS 2026 · similarity 0.82)
- MagicPIG: LSH Sampling for Efficient LLM Generation (ICLR 2025 · similarity 0.82)
- PiD: Fast and High-Resolution Latent Decoding with Pixel Diffusion (NeurIPS 2026 · similarity 0.82)
- Neural Physical Simulation with Multi-Resolution Hash Grid Encoding (AAAI 2024 · 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.
