Research map / Label-efficient and robust learning / Recognition and long-tail
Parametric Scattering Networks
Shanel Gauthier, Benjamin Thérien, Laurent Alsène-Racicot, Muawiz Chaudhary, Irina Rish, Eugene Belilovsky, Michael Eickenberg, Guy Wolf
Check 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 |
|---|---|---|---|
| Learn Morlet wavelet parameters from tight-frame initialization | Scattering + linear classifier · CIFAR-10, 100 samples | accuracy: 36.01±0.55 -> 37.84±0.57 | Improved |
| Learn Morlet wavelet parameters from tight-frame initialization | Scattering + WRN · COVIDx CRX-2, 100 samples | accuracy: 80.50± 1.15 -> 81.20±1.73 | Improved |
| Replace Morlet-constrained parameters with pixel-wise wavelet optimization (ablation) | Learned scattering + linear classifier · CIFAR-10, 100 samples | accuracy: 37.84±0.57 -> 32.30 ± 0.69 | Hurt |
| Use an equivariant parameterization with shared parameters across orientations (ablation) | Learned scattering + linear classifier · CIFAR-10, 100 samples | accuracy: 37.84±0.57 -> 39.69±0.56 | No clear effect |
| Use random rather than tight-frame initialization (ablation) | Learned scattering + linear classifier · CIFAR-10, 100 samples | accuracy: 37.84±0.57 -> 34.81±0.60 | Hurt |
| Increase learned filters per scale from 8 to 16 (ablation) | Learned scattering + linear classifier · CIFAR-10, 500 samples | accuracy: 50.74±0.30 -> 54.17±0.36 | Improved |
| Increase fixed scattering filters per scale from 8 to 16 (ablation) | Fixed scattering + linear classifier · CIFAR-10, 500 samples | accuracy: 49.67±0.33 -> 48.2 ± 0.27 | Hurt |
| Remove Autoaugment (ablation) | Learned scattering + linear classifier · CIFAR-10, all training data | accuracy: 69.57±0.1 -> 74.94±0.06 | Improved |
| Remove Autoaugment (ablation) | Learned scattering + WRN · CIFAR-10, 100 samples | accuracy: 43.60±0.87 -> 34.95 ± 0.96 | Hurt |
| Train wavelet parameters with SimCLR before linear evaluation | Scattering representation · CIFAR-10, 100 samples | accuracy: 36.01±0.55 -> 38.05 ± 0.45 | Improved |
| Replace cross-entropy with cosine loss (ablation) | Learned scattering + WRN · CIFAR-10 | accuracy: 43.6±0.87 -> 42.94 ± 0.77 | Hurt |
| Replace cross-entropy with cosine loss (ablation) | Learned scattering + WRN · KTH-TIPS2, 1188 samples | accuracy: 69.23±0.67 -> 70.86±0.67 | Improved |
Closest papers by meaning
- Kymatio: Scattering Transforms in Python (JMLR 2020 · similarity 0.89)
- Stability of Graph Scattering Transforms (NeurIPS 2019 · similarity 0.88)
- Diffusion Scattering Transforms on Graphs (ICLR 2019 · similarity 0.87)
- Unsupervised Parameter-free Simplicial Representation Learning with Scattering Transforms (ICML 2024 · similarity 0.87)
- Exploitation of a Latent Mechanism in Graph Contrastive Learning: Representation Scattering (NeurIPS 2024 · similarity 0.84)
- A General Graph Spectral Wavelet Convolution via Chebyshev Order Decomposition (ICML 2025 · similarity 0.83)
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.
