Research map / Architectures and efficiency / Neural architecture search
RF-Next: Efficient Receptive Field Search for Convolutional Neural Networks
Shanghua Gao, Zhongyu Li, Han Qi, Ming‐Ming Cheng, Liang Wang
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 |
|---|---|---|---|
| add EGI local search after global search (ablation) | MS-TCN · BreakFast | F@0.1: 72.2 -> 74.9 | Improved |
| add global search (ablation) | MS-TCN · BreakFast | F@0.1: 69.1 -> 72.2 | Improved |
| use EGI local search instead of DARTS (ablation) | MS-TCN · BreakFast | F@0.1: 73.4 -> 74.9 | Improved |
| add early stop to DARTS (ablation) | DARTS · BreakFast | F@0.1: 73.4 -> 73.8 | No clear effect |
| add Fair DARTS (ablation) | DARTS + early stop · BreakFast | F@0.1: 73.8 -> 73.3 | No clear effect |
| vary number of sampled dilation rates S (ablation) | EGI local search · BreakFast | F@0.1: 74.8 -> 74.9 | No clear effect |
| use proposed PMF instead of sigmoid (ablation) | EGI local search · BreakFast | F@0.1: 72.7 -> 74.9 | Improved |
| use proposed PMF instead of softmax (ablation) | EGI local search · BreakFast | F@0.1: 73.2 -> 74.9 | Improved |
| add local search on a random initial structure (ablation) | MS-TCN · BreakFast | F@0.1: 67.7 -> 73.6 | Improved |
| add local search on the human-designed baseline structure (ablation) | MS-TCN · BreakFast | F@0.1: 69.1 -> 74.1 | Improved |
| use structure searched on the same dataset (50Salads) instead of MS-TCN | MS-TCN · 50Salads | F@0.1: 67.1 -> 75.4 | Improved |
| use structure searched on 50Salads for GTEA | MS-TCN · GTEA | F@0.1: 83.8 -> 82.4 | Hurt |
Closest papers by meaning
- Receptive Fields As Experts in Convolutional Neural Architectures (ICML 2024 · similarity 0.87)
- Improving the Effective Receptive Field of Message-Passing Neural Networks (ICML 2025 · similarity 0.84)
- einspace: Searching for Neural Architectures from Fundamental Operations (NeurIPS 2024 · similarity 0.83)
- Your Image is My Video: Reshaping the Receptive Field via Image-To-Video Differentiable AutoAugmentation and Fusion (CVPR 2024 · similarity 0.83)
- Unveiling the Spatial-temporal Effective Receptive Fields of Spiking Neural Networks (NeurIPS 2025 · similarity 0.83)
- Neural Fields as Distributions: Signal Processing Beyond Euclidean Space (CVPR 2024 · similarity 0.82)
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.
