Research map / Architectures and efficiency / CNN design
Invertible and wide-network analysis
Builds invertible convolutional networks and analyzes infinitely wide networks through kernel theory.
65 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in CNN design, keyword cluster “convolution · kernel · spatial”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Quiet None of the tracked benchmarks has a report from 2025–26.
14 papers in 2023–24 and 6 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 10.2% to 5.9%.
No longer reported by papers here: CIFAR-10 (accuracy, last in 2023).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Multi-Bias Non-linear Activation in Deep Neural Networks (ICML 2016 · 46 citations)
- 2018 · Mixed Precision Training of Convolutional Neural Networks using Integer Operations (ICLR 2018 · 124 citations)
- 2019 · Making Convolutional Networks Shift-Invariant Again (ICML 2019 · 441 citations)
- 2020 · Fast Sparse ConvNets (CVPR 2020 · 104 citations)
- 2021 · Early Convolutions Help Transformers See Better (NeurIPS 2021 · 351 citations)
- 2022 · Feature Distillation Interaction Weighting Network for Lightweight Image Super-resolution (AAAI 2022 · 97 citations)
- 2023 · Learned Image Compression With Mixed Transformer-CNN Architectures (CVPR 2023 · 401 citations)
- 2024 · Convolutional Channel-Wise Competitive Learning for the Forward-Forward Algorithm (AAAI 2024 · 12 citations)
- 2025 · EfficientVMamba: Atrous Selective Scan for Light Weight Visual Mamba (AAAI 2025 · 165 citations)
- 2026 · Flowers: A Warp Drive for Neural PDE Solvers (ICML 2026 · 0 citations)
Earliest papers here
- Multi-Bias Non-linear Activation in Deep Neural Networks (ICML 2016 · 46 citations)
- Mixed Precision Training of Convolutional Neural Networks using Integer Operations (ICLR 2018 · 124 citations)
- A Deeper Look at Power Normalizations (CVPR 2018 · 60 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
CIFAR-10 · accuracy (higher is better)
6 papers, 2020–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 94.4 | Residual Distillation: Towards Portable Deep Neural Networks without Shortcuts (NeurIPS 2020 · 14 citations) |
What has been tried
59 papers here report 602 measured changes (a component added, removed or swapped, compared with a baseline): helped 66%, no clear effect 9%, hurt 21%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add ES2D by replacing SS2D | ImageNet · accuracy | 72.1 -> 73.6 | Helped | EfficientVMamba: Atrous Selective Scan for Light Weight Visual Mamba (AAAI 2025 · 165 citations) |
| use CwC loss instead of PvN loss | CIFAR-10 · test error (%) | 33.79 ± 0.84 -> 23.16 ± 0.16 | Helped | Convolutional Channel-Wise Competitive Learning for the Forward-Forward Algorithm (AAAI 2024 · 12 citations) |
| use convolutional tree depths 1,1,1,1 instead of the full depths | CIFAR-10 · accuracy | 84.99% -> 80.98% | Hurt | Convolutional Differentiable Logic Gate Networks (NeurIPS 2024 · 8 citations) |
| add time-frequency decoupling throughout the architecture | AudioSet full set, scratch · mAP | 0.468 -> 0.473 | Helped | DTF-AT: Decoupled Time-Frequency Audio Transformer for Event Classification (AAAI 2024 · 6 citations) |
| Replace final-stage FFNs with ShiftFFNs | ImageNet-1K · Top-1 accuracy | 82.119 -> 82.123 | No clear effect | FreeNet: Liberating Depth-Wise Separable Operations for Building Faster Mobile Vision Architectures (AAAI 2025 · 3 citations) |
| add a shift inside the reweighting function while leaving the multiplied feature unshifted | CIFAR-100, CF-ResNet-56 · Top-1 accuracy | 75.3˘0.4 -> 76.5˘0.3 | Helped | AdaShift: Learning Discriminative Self-Gated Neural Feature Activation With an Adaptive Shift Factor (CVPR 2024 · 2 citations) |
| repeat the downsampled image across 64 channels | ImageNette · accuracy | 57.8 ± 1.2 -> 56.3 ± 0.8 | Hurt | DEX: Data Channel Extension for Efficient CNN Inference on Tiny AI Accelerators (NeurIPS 2024 · 2 citations) |
| jointly optimize activation and convolution pruning instead of sequentially applying Depth then LayerOnly | ImageNet val · top-1 accuracy (%) | 71.72 -> 71.89 | Helped | LayerMerge: Neural Network Depth Compression through Layer Pruning and Merging (ICML 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 9 papers of this approach from 2024–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- CIFAR-10 · 3 of 9 papers
- ImageNet-1K · 3 of 9 papers
- MNIST · 2 of 9 papers
- CIFAR-100 · 2 of 9 papers
- ADE20K · 2 of 9 papers
- ImageNet · 2 of 9 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 20 papers; from OpenAlex and the papers' affiliations).
- University of Surrey · 2 papers
- Stanford University · 2 papers
Use it
Download BibTeX of 31 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Efficient CNN scaling and dynamic inference (104 papers)
- Dense matching and estimation CNNs (93 papers)
- Large and dynamic kernel design (62 papers)
- Higher-order pooling (20 papers)
Elsewhere in the topic
- Filter design and trained-filter analysis (128 papers)
- Compact CNN architectures (114 papers)
- Theory of convolutional inductive bias (64 papers)
- Group symmetry equivariant networks (55 papers)
- Biologically inspired CNN explanation (46 papers)
- CNN filter pruning (43 papers)
- Steerable rotation-equivariant CNNs (40 papers)
- Group convolution theory (19 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).