Research map / Graph learning / GNN architectures
Generalised Implicit Neural Representations
Daniele Grattarola, Pierre Vandergheynst
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 spectral embeddings of the graph as INR input instead of node coordinates (ReLU activation) | INR · Bunny held-out R^2 | R^2: 0.000 -> 0.932 | Improved |
| Use spectral embeddings instead of node coordinates (ReLU activation) | INR · Protein held-out R^2 | R^2: 0.799 -> 0.921 | Improved |
| Use spectral embeddings instead of node coordinates (SIREN activation) | INR · Protein held-out R^2 | R^2: 0.275 -> 0.916 | Improved |
| Use spectral embeddings instead of node coordinates (SIREN activation) | INR · Bunny held-out R^2 | R^2: 0.919 -> 0.885 | Hurt |
| Reduce spectral embedding size to k<=3 (ablation) | Generalised INR · Bunny | R^2 | Hurt |
| Use large spectral embedding size k=100 (ablation) | Generalised INR · Bunny | R^2 | Improved |
| Use smaller k and ReLU activations to improve transfer to new graph realisations | Generalised INR · Transferability | transfer quality | Improved |
| Increase k in super-resolution transfer (ablation) | Generalised INR · Bunny super-resolution | R^2 | Mixed |
| Ignore nodes above 90th percentile of squared error in super-resolution evaluation | Generalised INR · Bunny super-resolution | R^2: 0.39 -> 0.94 | Improved |
| Larger latent size z in multi-protein autodecoder (ablation) | Conditional GINR · Multi-protein INR | R^2 | No clear effect |
| Increase number of proteins in the dataset (ablation) | Conditional GINR · Multi-protein INR | R^2 | Hurt |
| Use different Laplacian normalisation for spectral embeddings (ablation) | Generalised INR · Appendix | performance | No clear effect |
Closest papers by meaning
- A sampling theory perspective on activations for implicit neural representations (ICML 2024 · similarity 0.87)
- Grids Often Outperform Implicit Neural Representation at Compressing Dense Signals (NeurIPS 2025 · similarity 0.87)
- Implicit Neural Representations and the Algebra of Complex Wavelets (ICLR 2024 · similarity 0.87)
- Neural Fields as Distributions: Signal Processing Beyond Euclidean Space (CVPR 2024 · similarity 0.85)
- GIST: Gauge-Invariant Spectral Transformers for Scalable Graph Neural Operators (NeurIPS 2026 · similarity 0.85)
- Consistent Geometric Deep Learning via Hilbert Bundles and Cellular Sheaves (NeurIPS 2026 · similarity 0.85)
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.
