Research map / Architectures and efficiency / RNNs and normalization
Beyond Periodicity: Towards a Unifying Framework for Activations in Coordinate-MLPs
Sameera Ramasinghe, Simon Lucey
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 |
|---|---|---|---|
| use Gaussian activation instead of ReLU | coordinate-MLP · 2D signal encoding, natural images | PSNR (dB): 20.42 -> 33.43 | Improved |
| use Gaussian activation instead of ReLU | coordinate-MLP · 2D signal encoding, text images | PSNR (dB): 18.49 -> 36.17 | Improved |
| use Gaussian activation instead of ReLU | coordinate-MLP · 2D signal encoding, noise images | PSNR (dB): 10.82 -> 11.78 | Improved |
| use Laplacian activation instead of SiLU | coordinate-MLP · 2D signal encoding, natural images | PSNR (dB): 17.59 -> 33.01 | Improved |
| use ExpSin non-periodic activation instead of Tanh | coordinate-MLP · 2D signal encoding, text images | PSNR (dB): 16.19 -> 35.89 | Improved |
| use Laplacian activation instead of SoftPlus | coordinate-MLP · 2D signal encoding, noise images | PSNR (dB): 9.71 -> 11.01 | Mixed |
| use Gaussian activations without positional embeddings | coordinate-MLP · Novel view synthesis | visual fidelity | Improved |
| use Gaussian activation with positional embedding in shallower MLP | coordinate-MLP · Novel view synthesis | parameters / quality: ~50% less parameters | Improved |
| use Gaussian instead of sine under non-SIREN initialization | coordinate-MLP · Image fitting convergence | convergence | Improved |
Closest papers by meaning
- Neural Functions for Learning Periodic Signal (ICLR 2025 · similarity 0.87)
- Coordinate-Aware Modulation for Neural Fields (ICLR 2024 · similarity 0.86)
- A sampling theory perspective on activations for implicit neural representations (ICML 2024 · similarity 0.85)
- Representing Sounds as Neural Amplitude Fields: A Benchmark of Coordinate-MLPs and a Fourier Kolmogorov-Arnold Framework (AAAI 2025 · similarity 0.85)
- ASMR: Activation-Sharing Multi-Resolution Coordinate Networks for Efficient Inference (ICLR 2024 · similarity 0.84)
- Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion (ICLR 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.
