Research map / Probabilistic and scientific modeling / Variational and generative sampling
Modality-Agnostic Variational Compression of Implicit Neural Representations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 low-rank soft-gating masks predicted nonlinearly from latent modulations instead of sparse MSCN modulations | INR conditioning · ERA5 (4×), dim 1024 | PSNR: more than 3.1dB | Helped |
| Use VC-INR conditioning instead of MSCN conditioning | INR conditioning · CelebA-HQ, dim 1024 | PSNR | Hurt |
| Use VC-INR conditioning instead of MSCN conditioning | INR conditioning · SRN Cars, dim 128 and 256 | PSNR | No clear effect |
| Replace uniform quantisation with learned variational transform coding | VC-INR latent compression · CIFAR-10 | PSNR at a given bitrate | Helped |
| Use LayerNorm in the latent-to-mask predictor (ablation) | VC-INR meta-learning · Meta-training | training stability | Helped |
| Increase the width of the nonlinear latent-to-mask predictor (ablation) | VC-INR meta-learning · Pre-training | reconstruction performance | Helped |
| Use nonlinear mappings from latent parameters to modulations rather than Functa's linear projection | INR conditioning · Multiple datasets | reconstruction performance | Helped |
| Use nonlinear mappings from latent parameters to modulations rather than Functa's linear projection | INR conditioning · Multiple datasets | learning progress | Helped |
| Increase latent modulation size (ablation) | VC-INR · ERA5 (4×) | PSNR: 45.0, 46.2, 47.6, 49.0, 50.0 | Helped |
| Increase latent modulation size (ablation) | VC-INR · CelebA-HQ | PSNR: 22.0, 23.9, 26.0, 28.3, 30.8 | Helped |
| Increase latent modulation size (ablation) | VC-INR · SRN Cars | PSNR: 23.9, 24.0, 24.3, 25.2, 25.5 | Helped |
| Use data-space reconstruction distortion rather than latent-space MSE as the compression distortion (ablation) | VC-INR quantiser training · Compression experiments | reconstruction quality | Helped |
Closest papers by meaning
- SINR: Sparsity Driven Compressed Implicit Neural Representations (CVPR 2025 · similarity 0.89)
- NVRC: Neural Video Representation Compression (NeurIPS 2024 · similarity 0.89)
- Towards Lossless Implicit Neural Representation via Bit Plane Decomposition (CVPR 2025 · similarity 0.88)
- On Quantizing Neural Representation for Variable-Rate Video Coding (ICLR 2025 · similarity 0.87)
- Robustly overfitting latents for flexible neural image compression (NeurIPS 2024 · similarity 0.86)
- VI^3NR: Variance Informed Initialization for Implicit Neural Representations (CVPR 2025 · similarity 0.86)
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.