Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Bayesian Deep Convolutional Networks with Many Channels are 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 global average pooling | CNN-GP on CIFAR10 · CIFAR10 | test error (%): 32.86 -> 22.57 | Helped |
| Use global average pooling | SGD-trained CNN on CIFAR10 · CIFAR10 | test error (%): 24.76 -> 14.85 | Helped |
| Replace the infinite-width Bayesian CNN-GP with a carefully tuned finite CNN trained using SGD | CNN without pooling on CIFAR10 · CIFAR10 | test error (%): 32.86 -> 24.76 | Helped |
| Allow SGD training to underfit rather than requiring 100% training accuracy | CNN with ReLU and large learning rate on CIFAR10 · CIFAR10 | test error (%): 24.76 (17.64) | Helped |
| Increase channel count | finite CNNs and LCNs · CIFAR10 | validation accuracy | Helped |
| Estimate the GP kernel with more channels and more Monte Carlo samples | MC-CNN-GP · CIFAR10 downsampled to 8×8 | validation accuracy | Helped |
Closest papers by meaning
- Deep Neural Networks as Gaussian Processes (ICLR 2018 · similarity 0.90)
- Approximate Inference Turns Deep Networks into Gaussian Processes (NeurIPS 2019 · similarity 0.88)
- Gaussian Process Behaviour in Wide Deep Neural Networks (ICLR 2018 · similarity 0.88)
- Exact Gaussian Moment Matching for Residual Networks: a Second-Order Method (NeurIPS 2026 · similarity 0.86)
- Partially Stochastic Infinitely Deep Bayesian Neural Networks (ICML 2024 · similarity 0.86)
- Bayesian Optimization via Continual Variational Last Layer Training (ICLR 2025 · 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.