Research map / Architectures and efficiency / CNN design
Sparsely Aggregated Convolutional Networks
Ligeng Zhu, Ruizhi Deng, Michael Maire, Zhiwei Deng, Greg Mori, Ping Tan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 dense aggregation with sparse (logarithmic) aggregation in DenseNet-BC 100 k=32/64/128 | DenseNet-BC · CIFAR-100+ | error (%): 18.71 -> 18.22 | Improved |
| replace dense aggregation with sparse aggregation in DenseNet-BC (parameters) | DenseNet-BC · CIFAR-100+ | parameters: 30.5M -> 16.7M | Improved |
| replace dense aggregation with sparse aggregation in DenseNet-40 k=12 | DenseNet · CIFAR-100+ | error (%): 24.79 -> 24.48 | Improved |
| replace dense with sparse aggregation in DenseNet-100 k=12 (similar 7.2M params) | DenseNet · CIFAR-100+ | error (%): 20.97 -> 19.89 | Improved |
| replace dense with sparse aggregation in DenseNet-BC-250 k=24 | DenseNet-BC · CIFAR-100+ | error (%): 17.6 -> 18.22 | Hurt |
| replace dense with sparse aggregation in DenseNet-169-32 on ImageNet | DenseNet · ImageNet val | top-1 error (%): 23.6 -> 23.6 | Mixed |
| replace dense with sparse aggregation in DenseNet-201-32 on ImageNet | DenseNet · ImageNet val | top-1 error (%): 22.6 -> 22.7 | Mixed |
| replace dense with sparse aggregation in DenseNet-121-32 on ImageNet (FLOPs) | DenseNet · ImageNet val | FLOPs: 5.7G -> 3.46G | Improved |
| scale depth from 1001 to 2000 for ResNet | ResNet · CIFAR-100+ | error (%): 21.42 -> 22.76 | Hurt |
| scale depth from 1001 to 2000 for sparse-sum SparseNet | SparseNet[+] · CIFAR-100+ | error (%): 22.10 -> 21.01 | Improved |
| replace summation with concatenation in sparse aggregation (SparseNet[concat]-400 vs SparseNet[+]-200) | SparseNet · CIFAR-100+ | error (%): 25.77 -> 24.53 | Improved |
| use sparse instead of dense aggregation at extreme depth (DenseNet-BC k=4 1001 vs SparseNet-BC k=12 400) | DenseNet-BC · CIFAR-100+ | error (%): 28.50 -> 24.53 | Improved |
Closest papers by meaning
- Deep Layer Aggregation (CVPR 2018 · similarity 0.90)
- Fast Sparse ConvNets (CVPR 2020 · similarity 0.88)
- Inference, Learning and Attention Mechanisms that Exploit and Preserve Sparsity in CNNs (IJCV 2020 · similarity 0.88)
- Sparse Cocktail: Every Sparse Pattern Every Sparse Ratio All At Once (ICML 2024 · similarity 0.87)
- LAuReL: Learned Augmented Residual Layer (ICML 2025 · similarity 0.86)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (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.
