Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Learning from Small Sample Sets by Combining Unsupervised Meta-Training with CNNs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 unsupervised LDS meta-training layer on top of pre-trained CNN (single-scale) | VGG19 fc7 features · SUN-397, MIT-67, Flowers, Stanford-40 (full training sets) | classification accuracy (%) | Helped |
| multi-scale LDS layers across several top layers | VGG19 pre-trained CNN · SUN-397, MIT-67, Flowers, Stanford-40 | classification accuracy (%) | Helped |
| single-scale LDS meta-training for one-shot learning on MIT-67 | VGG19 off-the-shelf features · MIT-67 one-shot | accuracy gain over pre-trained CNN: nearly 20% | Helped |
| fine-tune LDS+CNN instead of using off-the-shelf features (AlexNet) | single-scale and multi-scale LDS+CNN · SUN-397, MIT-67 | classification accuracy (%) | Helped |
| LDS meta-training on unlabeled Flickr YFCC100M instead of weakly-supervised caption training from scratch | CNN features · SUN-397, MIT-67, Flowers, Stanford-40 | classification accuracy (%) | Helped |
| reduce LDS+CNN dimensionality from 20,000 to 4,096 via RFE vs PCA (ablation) | single-scale LDS+CNN · few-shot scene/flower/action benchmarks | classification accuracy | Mixed |
Closest papers by meaning
- Unsupervised Deep Learning by Neighbourhood Discovery (ICML 2019 · similarity 0.88)
- A Dual Attention Network with Semantic Embedding for Few-Shot Learning (AAAI 2019 · similarity 0.88)
- Unsupervised Deep Learning via Affinity Diffusion (AAAI 2020 · similarity 0.87)
- Unsupervised Meta-Learning for Few-Shot Image Classification (NeurIPS 2019 · similarity 0.87)
- Unsupervised Learning via Meta-Learning (ICLR 2019 · similarity 0.87)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 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.