Research map / Architectures and efficiency / CNN design
Deep SimNets
Check 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 |
|---|---|---|---|
| Replace the Caffe ConvNet with a two layer SimNet | Caffe ConvNet · CIFAR-10 test | accuracy (%): 81.1 vs 85.5 | Helped |
| Replace the Caffe ConvNet with a two layer SimNet | Caffe ConvNet · SVHN test | accuracy (%): 94 vs 93.8 | No clear effect |
| Replace the Caffe ConvNet with a two layer SimNet | Caffe ConvNet · CIFAR-100 test | accuracy (%): 52.4 vs 54.6 | Helped |
| Use a three layer SimNet instead of the listed ConvNet baselines | CIFAR-10 image classification · CIFAR-10 | accuracy (%): 92.18 vs 91.19; 92.03; 92.4; 92.75 | Mixed |
| Reduce every layer to 10 channels and remove multiplicative Gaussian noise (ablation) | Three layer SimNet · CIFAR-10 | accuracy (%): 92.18 vs 76.8% | Hurt |
| Use a compact three layer SimNet instead of compact Network in Network (ablation) | Network in Network · CIFAR-10 | accuracy (%): 72.3% vs 76.8% | Helped |
| Use weighted ℓp similarity instead of linear similarity in a single-layer network (ablation) | Single-layer ConvNet · CIFAR-10 cross-validation | accuracy | Helped |
| Use ℓ1 rather than ℓ2 similarity in a single-layer SimNet (ablation) | Single-layer SimNet · CIFAR-10 cross-validation | accuracy | No clear effect |
| Use a two layer SimNet rather than Caffe ConvNet at the reported model sizes | Caffe ConvNet · CIFAR-10 | FLOPs required to classify an image: 24.8M vs 14.2M | Helped |
Closest papers by meaning
- Approximation and Learning with Deep Convolutional Models: a Kernel Perspective (ICLR 2022 · similarity 0.87)
- Fast Sparse ConvNets (CVPR 2020 · similarity 0.86)
- EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks (ICML 2019 · similarity 0.86)
- Linear CNNs Discover the Statistical Structure of the Dataset Using Only the Most Dominant Frequencies (ICML 2023 · similarity 0.86)
- Boosting Dilated Convolutional Networks with Mixed Tensor Decompositions (ICLR 2018 · similarity 0.86)
- Unveiling the Unseen: Identifiable Clusters in Trained Depthwise Convolutional Kernels (ICLR 2024 · 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.