Research map / Label-efficient and robust learning / Domain adaptation and generalization
Adapted Deep Embeddings: A Synthesis of Methods for k-Shot Inductive Transfer 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 |
|---|---|---|---|
| Adapt the histogram-loss embedding using target support data | Histogram-loss embedding · Omniglot, restricted-source experiments (k > 1) | accuracy: mean 37.0% | Helped |
| Adapt the prototypical embedding using target support data | Prototypical network · MNIST (k > 1) | accuracy | Helped |
| Adapt the histogram-loss embedding using target support data | Histogram-loss embedding · MNIST (k > 1) | accuracy | Helped |
| Adapt the embedding using target support data | Histogram-loss and prototypical embeddings · MNIST (k = 1) | accuracy | No clear effect |
| Use histogram loss rather than prototypical-network training | Adapted embeddings · MNIST, Isolet, tinyImageNet, and Omniglot | accuracy | Helped |
| Use episodic training rather than restricted-source training | Omniglot embedding methods · Omniglot | accuracy | Mixed |
| Add target-domain weight adaptation to source-trained embeddings | Adapted embeddings · 34 dataset, k, and n conditions | classification error reduction: 33.7% | Helped |
| Use deep embeddings instead of weight adaptation | k-shot inductive transfer · MNIST, Isolet, tinyImageNet, and Omniglot | accuracy | Helped |
| Transfer source-trained weights rather than train only on target data | Weight adaptation · Omniglot (n > 5) | accuracy | No clear effect |
| Increase k beyond 50 without adapting the prototypical embedding | Prototypical network · MNIST | accuracy | No clear effect |
| Adapt the prototypical embedding as more target examples become available | Adapted prototypical network · MNIST | accuracy | Helped |
| Add episodic training to the Omniglot comparison (ablation) | Histogram-loss and prototypical embeddings · Omniglot (5, 20) | accuracy: 0.9839 ± 0.0011 -> 0.9797 ± 0.0014; 0.9864 ± 0.0009 -> 0.9847 ± 0.0011 | Mixed |
Closest papers by meaning
- Embedding Transfer With Label Relaxation for Improved Metric Learning (CVPR 2021 · similarity 0.88)
- Domain Adaptation by Mixture of Alignments of Second- or Higher-Order Scatter Tensors (CVPR 2017 · similarity 0.88)
- Few-Shot Learning via Embedding Adaptation With Set-to-Set Functions (CVPR 2020 · similarity 0.87)
- A General Representation-Based Approach to Multi-Source Domain Adaptation (ICML 2025 · similarity 0.87)
- HiGDA: Hierarchical Graph of Nodes to Learn Local-to-Global Topology for Semi-Supervised Domain Adaptation (AAAI 2025 · similarity 0.86)
- Flatten Long-Range Loss Landscapes for Cross-Domain Few-Shot Learning (CVPR 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.