Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Unsupervised Learning via Meta-Learning
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 |
|---|---|---|---|
| Meta-learn on tasks generated by clustering ACAI embeddings rather than directly fitting a classifier on the embeddings | ACAI representations · Omniglot | accuracy: 61.08% -> 68.84% | Helped |
| Meta-learn on tasks generated by clustering DeepCluster embeddings rather than directly fitting a classifier on the embeddings | DeepCluster representations · miniImageNet | accuracy: 29.44% -> 39.90% | Helped |
| Meta-learn from clustered BiGAN embeddings for attribute tasks | BiGAN representations · CelebA | accuracy: 63.19% -> 74.98% | Helped |
| Use multiple clustering partitions instead of one (ablation) | ACAI CACTUs-MAML · Omniglot (5-way, 1-shot) | accuracy: 66.49% -> 68.84% | Helped |
| Use random image-to-cluster assignments instead of structured embedding clusters (ablation) | Omniglot MAML · Omniglot (5-way, 1-shot) | accuracy: 52.50% -> 25.99% | Hurt |
| Cluster in pixel space rather than a learned embedding (ablation) | Omniglot MAML · Omniglot (5-way, 1-shot) | accuracy: 52.50% -> 30.55% | Hurt |
| Replace embedding clusters with zero-margin random hyperplane partitions (ablation) | BiGAN MAML · miniImageNet (5-way, 1-shot) | accuracy: 37.75% -> 20.00% | Hurt |
| Add a nonzero margin to random hyperplane partitions (ablation) | BiGAN Hyperplanes-MAML · miniImageNet (5-way, 1-shot) | accuracy: 20.00% -> 29.67% | Helped |
| Use multiple partitions instead of one (ablation) | DeepCluster CACTUs-MAML · miniImageNet (5-way, 1-shot) | accuracy: 38.75% -> 39.90% | Helped |
Closest papers by meaning
- Unsupervised Meta-Learning via In-Context Learning (ICLR 2025 · similarity 0.92)
- Meta-GMVAE: Mixture of Gaussian VAE for Unsupervised Meta-Learning (ICLR 2021 · similarity 0.90)
- Unsupervised Meta-Learning for Few-Shot Image Classification (NeurIPS 2019 · similarity 0.90)
- Meta-Learning without Memorization (ICLR 2020 · similarity 0.90)
- Unsupervised Meta-learning via Few-shot Pseudo-supervised Contrastive Learning (ICLR 2023 · similarity 0.90)
- On the Provable Advantage of Unsupervised Pretraining (ICLR 2024 · similarity 0.89)
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.