Research map / Architectures and efficiency / CNN design
Deep Learning for Fixed Model Reuse
Yang Yang, De-Chuan Zhan, Ying Fan, Yuan Jiang, Zhi-Hua Zhou
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 |
|---|---|---|---|
| parallel structure with knockdown instead of sequential target replacing (SQR) (ablation) | FMR · WIKI | accuracy (%): 42.7 -> 43.3 | Improved |
| parallel structure with knockdown instead of SQR (ablation) | FMR · Flickr8k | accuracy (%): 72.6 -> 74.8 | Improved |
| parallel structure with knockdown instead of SQR (ablation) | FMR · HMDB51 | accuracy (%): 62.9 -> 68.9 | Improved |
| parallel structure with knockdown instead of SQR (ablation) | FMR · UCF101 | accuracy (%): 85.3 -> 91.6 | Improved |
| parallel structure with knockdown instead of SQR (ablation) | FMR · UCF50 | accuracy (%): 91.1 -> 92.4 | Improved |
| double features dropped per iteration from 52 to 104 (ablation) | FMR · UCF101 | accuracy (%): 91.6 -> 92.0 | No clear effect |
| add FMR (reusing fixed text features) to CNN trained on images only | CNN · WIKI | accuracy (%): 42.6 -> 43.3 | Improved |
| use FMR instead of MIFS fixed features at test time | FMR · UCF50 | accuracy (%): 94.4 -> 92.4 | Hurt |
| add knockdown to FMR training (ablation) | CNN · UCF101 | accuracy (%) | Improved |
Closest papers by meaning
- Model Reprogramming: Resource-Efficient Cross-Domain Machine Learning (AAAI 2024 · similarity 0.84)
- EMR-Merging: Tuning-Free High-Performance Model Merging (NeurIPS 2024 · similarity 0.83)
- Update Your Transformer to the Latest Release: Re-Basin of Task Vectors (ICML 2025 · similarity 0.82)
- RESQUE: Quantifying Estimator to Task and Distribution Shift for Sustainable Model Reusability (AAAI 2025 · similarity 0.82)
- Inductive biases of multi-task learning and finetuning: multiple regimes of feature reuse (NeurIPS 2024 · similarity 0.81)
- Reparameterized Multi-Resolution Convolutions for Long Sequence Modelling (NeurIPS 2024 · similarity 0.81)
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.
