Research map / Architectures and efficiency / CNN design
End-to-End Kernel Learning with Supervised Convolutional Kernel Networks
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 supervised CKN instead of Gen P. | Image classification · CIFAR-10 test | test error (%): 7.62 -> 10.20 | Hurt |
| Use supervised CKN instead of Gen P. | Image classification · SVHN test | test error (%): 1.69 -> 2.04 | Hurt |
| Use SCKN instead of CSCN. | Super-resolution · Set5 x2 | PSNR: 36.93 -> 37.07 | Helped |
| Use SCKN instead of CSCN. | Super-resolution · Set14 x2 | PSNR: 32.56 -> 32.76 | Helped |
| Use SCKN instead of CSCN. | Super-resolution · Kodim x2 | PSNR: 32.94 -> 33.21 | Helped |
| Use SCKN instead of CSCN. | Super-resolution · Set5 x3 | PSNR: 33.10 -> 33.08 | Hurt |
| Use SCKN instead of CSCN. | Super-resolution · Set14 x3 | PSNR: 29.41 -> 29.50 | Helped |
| Use SCKN instead of CSCN. | Super-resolution · Kodim x3 | PSNR: 29.76 -> 29.88 | Helped |
| Replace the original CKN with the improved unsupervised SCKN approach. | CIFAR-10 image classification · CIFAR-10 test | test error (%): 21.7% -> 19.3% | Helped |
| Reduce filters per layer from 512 to 128. (ablation) | SCKN · CIFAR-10 test | test error (%): 10.20 -> 11.95% | Hurt |
| Reduce filters per layer from 512 to 128. (ablation) | SCKN · SVHN test | test error (%): 2.04 -> 2.15% | Hurt |
| Use the paper's unsupervised two-layer network with (1024-16384) filters. (ablation) | SCKN · CIFAR-10 test | test error (%): 10.20 -> 14.2% | Mixed |
Closest papers by meaning
- The Unreasonable Effectiveness of Patches in Deep Convolutional Kernels Methods (ICLR 2021 · similarity 0.91)
- Convolutional Spectral Kernel Learning with Generalization Guarantees (Abstract Reprint) (AAAI 2024 · similarity 0.90)
- Stochastic Kernel Regularisation Improves Generalisation in Deep Kernel Machines (NeurIPS 2024 · similarity 0.88)
- Unveiling the Unseen: Identifiable Clusters in Trained Depthwise Convolutional Kernels (ICLR 2024 · similarity 0.88)
- KernelWarehouse: Rethinking the Design of Dynamic Convolution (ICML 2024 · similarity 0.88)
- Convolutional Deep Kernel Machines (ICLR 2024 · similarity 0.87)
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.