Research map / Architectures and efficiency / CNN design
Towards Biologically Plausible Convolutional 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 |
|---|---|---|---|
| Add dynamic weight sharing every batch (ablation) | Locally connected ResNet20 · CIFAR10 (padding 4) | Top-1 accuracy (%): 80.9 -> 87.4 | Helped |
| Add 16 translation repetitions (ablation) | Locally connected ResNet20 · CIFAR10 (padding 4) | Top-1 accuracy (%): 80.9 -> 85.0 | Helped |
| Use dynamic weight sharing every 10 batches rather than every batch (ablation) | Locally connected ResNet20 · CIFAR10 (padding 4) | Top-1 accuracy (%): 87.4 -> 85.1 | Hurt |
| Use dynamic weight sharing every 100 batches rather than every batch (ablation) | Locally connected ResNet20 · CIFAR10 (padding 4) | Top-1 accuracy (%): 87.4 -> 82.0 | Hurt |
| Add dynamic weight sharing every batch (ablation) | Locally connected ResNet20 · CIFAR100 (padding 4) | Top-1 accuracy (%): 49.8 -> 58.7 | Helped |
| Add 16 translation repetitions (ablation) | Locally connected ResNet20 · CIFAR100 (padding 4) | Top-1 accuracy (%): 49.8 -> 55.9 | Helped |
| Add dynamic weight sharing every batch (ablation) | Locally connected ResNet20 · TinyImageNet (padding 4) | Top-1 accuracy (%): 29.6 -> 41.6 | Helped |
| Add 16 translation repetitions (ablation) | Locally connected ResNet20 · TinyImageNet (padding 4) | Top-1 accuracy (%): 29.6 -> 34.0 | Helped |
| Add dynamic weight sharing every 10 batches (ablation) | Locally connected ResNet18 (0.5x width) · ImageNet | Top-1 accuracy (%): 46.7 -> 59.3 | Helped |
| Reduce dynamic weight-sharing frequency from every 1 to every 10 batches (ablation) | Locally connected ResNet18 (0.5x width) · ImageNet | Top-1 accuracy (%): 61.7 -> 59.3 | Hurt |
| Add dynamic weight sharing every batch (ablation) | Locally connected ResNet18 (0.5x width) · Brain-Score (ImageNet-trained networks) | average score: .349 -> .396 | Helped |
| Add dynamic weight sharing every 10 batches (ablation) | Locally connected ResNet18 (0.5x width) · Brain-Score (ImageNet-trained networks) | average score: .349 -> .385 | Helped |
Closest papers by meaning
- Explicitly Modeling Subcortical Vision with a Neuro-Inspired Front-End Improves CNN Robustness (NeurIPS 2025 · similarity 0.86)
- Role of Locality and Weight Sharing in Image-Based Tasks: A Sample Complexity Separation between CNNs, LCNs, and FCNs (ICLR 2024 · similarity 0.86)
- Computational Separation Between Convolutional and Fully-Connected Networks (ICLR 2021 · similarity 0.86)
- Structured and Deep Similarity Matching via Structured and Deep Hebbian Networks (NeurIPS 2019 · similarity 0.86)
- Local plasticity rules can learn deep representations using self-supervised contrastive predictions (NeurIPS 2021 · similarity 0.86)
- Dendritic Localized Learning: Toward Biologically Plausible Algorithm (ICML 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.