Research map / Architectures and efficiency / CNN design
Deep Convolutional Networks as shallow Gaussian Processes
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 the proposed ResNet GP kernel rather than the NNGP | MNIST classification · MNIST test | classification error: 1.21% -> 0.84% | Helped |
| Use the proposed ResNet GP kernel rather than the Convolutional GP | MNIST classification · MNIST test | classification error: 1.17% -> 0.84% | Helped |
| Use the proposed ResNet GP kernel rather than the Deep Conv. GP | MNIST classification · MNIST test | classification error: 1.34% -> 0.84% | Helped |
| Use the proposed ResNet GP kernel rather than the parametric deep kernel | MNIST classification · MNIST test | classification error: 0.60% -> 0.84% | Hurt |
| Use the proposed ResNet GP rather than a standard ResNet | MNIST classification · MNIST test | classification error: 0.41% -> 0.84% | Hurt |
| Add residual connections to the convolutional GP | ConvNet GP · MNIST test | classification error: 1.03% -> 0.93% | Helped |
| Use a 32-layer ResNet architecture rather than the random residual CNN architecture | Residual CNN GP · MNIST test | classification error: 0.93% -> 0.84% | Helped |
| Use the convolutional kernel rather than an RBF kernel with 1000 inducing points | GP classification with RobustMax likelihood · MNIST test, inducing-point setting | classification error: 3.4% -> 2.4% | Helped |
Closest papers by meaning
- End-to-End Kernel Learning with Supervised Convolutional Kernel Networks (NeurIPS 2016 · similarity 0.89)
- Approximation and Learning with Deep Convolutional Models: a Kernel Perspective (ICLR 2022 · similarity 0.89)
- CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters (CVPR 2022 · similarity 0.88)
- Convolutional Deep Kernel Machines (ICLR 2024 · similarity 0.87)
- Stochastic Kernel Regularisation Improves Generalisation in Deep Kernel Machines (NeurIPS 2024 · similarity 0.86)
- Unveiling the Unseen: Identifiable Clusters in Trained Depthwise Convolutional Kernels (ICLR 2024 · 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.