Research map / Architectures and efficiency / Pruning and efficient tuning
RIFLE: Backpropagation in Depth for Deep Transfer Learning through Re-Initializing the Fully-connected LayEr
Xingjian Li, Haoyi Xiong, Haozhe An, Cheng-Zhong Xu, Dejing Dou
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 |
|---|---|---|---|
| Periodically reinitialize the FC layer and use cyclic learning rates | L2 fine-tuning · Image classification transfer benchmarks (8 datasets) | top-1 accuracy | Mixed |
| Periodically reinitialize the FC layer and use cyclic learning rates | L2-SP fine-tuning · Image classification transfer benchmarks (8 datasets) | top-1 accuracy | Improved |
| Remove cyclic learning rate (retain FC reinitialization) (ablation) | RIFLE + L2 · Flower-102 | top-1 accuracy: 93.00$\pm$0.08 -> 92.17$\pm$0.19 | Hurt |
| Remove cyclic learning rate (retain FC reinitialization) (ablation) | RIFLE + L2 · Food-101-150 | top-1 accuracy: 76.36$\pm$0.16 -> 76.66$\pm$0.11 | Improved |
| Remove FC reinitialization (retain cyclic learning rate) (ablation) | RIFLE + L2-SP · Food-101-150 | top-1 accuracy: 77.36$\pm$0.04 -> 76.44$\pm$0.11 | Hurt |
| Remove FC reinitialization (retain cyclic learning rate) (ablation) | RIFLE + L2 · Flower-102 | top-1 accuracy: 93.00$\pm$0.08 -> 92.44$\pm$0.16 | Hurt |
Closest papers by meaning
- Auto-Transfer: Learning to Route Transferable Representations (ICLR 2022 · similarity 0.85)
- Towards Inadequately Pre-trained Models in Transfer Learning (ICCV 2023 · similarity 0.85)
- Reducing Fine-Tuning Memory Overhead by Approximate and Memory-Sharing Backpropagation (ICML 2024 · similarity 0.83)
- Accelerated training through iterative gradient propagation along the residual path (ICLR 2025 · similarity 0.83)
- LoKi: Low-dimensional KAN for Efficient Fine-tuning Image Models (CVPR 2025 · similarity 0.83)
- Features are fate: a theory of transfer learning in high-dimensional regression (ICML 2025 · similarity 0.83)
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.
