Research map / Label-efficient and robust learning / Domain adaptation and generalization
d-SNE: Domain Adaptation Using Stochastic Neighborhood Embedding
Xiang Xu, Xiong Zhou, Ragav Venkatesan, Gurumurthy Swaminathan, Orchid Majumder
Check 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 |
|---|---|---|---|
| Use d-SNE instead of FADA with one labeled target example per class | MNIST → USPS · MNIST → USPS, 1 target sample/class | accuracy: 89.10 -> 92.90 | Improved |
| Use d-SNE instead of CCSA with no labeled target examples | MNIST → USPS · MNIST → USPS, 0 target samples/class | accuracy: 65.40 -> 73.01 | Improved |
| Use d-SNE instead of CCSA with 10 labeled target examples per class | MNIST → USPS · MNIST → USPS | accuracy: 97.27 ± 0.19 -> 99.00 ± 0.08 | Improved |
| Use d-SNE instead of the strongest listed unsupervised result | MNIST ↔ USPS · MNIST → USPS | accuracy: 98.26 ± 0.11 -> 99.00 ± 0.08 | Improved |
| Use supervised d-SNE instead of DIRT-T | SVHN → MNIST · SVHN → MNIST | accuracy: 99.40 -> 96.45 ± 0.20 | Hurt |
| Add the semi-supervised extension to supervised d-SNE | MNIST → MNIST-M · MNIST → MNIST-M | accuracy: 87.80 ± 0.16 -> 94.12 | Improved |
| Add the semi-supervised extension to supervised d-SNE | MNIST → SVHN · MNIST → SVHN | accuracy: 61.73 ± 0.47 -> 77.63 ± 0.26 | Improved |
| Adapt the model to a target domain, then evaluate on the source domain | d-SNE · MNIST → MNIST-M, source accuracy | accuracy: 99.45% -> 99.51% | Improved |
| Adapt the model to a target domain, then evaluate on the source domain | d-SNE · MNIST → SVHN, source accuracy | accuracy: 99.51% -> 99.59% | Improved |
| Adapt the model to a target domain, then evaluate on the source domain | d-SNE · SVHN → MNIST, source accuracy | accuracy: 88.96% -> 94.94% | Improved |
| Use d-SNE with ResNet-101 instead of CCSA with VGG-16 | Office-31 · Office-31 average across six transfers, 3 target samples/class | accuracy: 85.80 -> 90.01 | Improved |
| Use d-SNE instead of CCSA | VisDA-C · VisDA-C, synthetic → real, supervised | accuracy: 76.89 -> 80.66 | Improved |
Later papers that compared with this method
2 later accepted papers report d-SNE as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Augmenting and Aligning Snippets for Few-Shot Video Domain Adaptation (ICCV 2023 · better on 6 of 6 shared results)
- Out-of-Domain Generalization From a Single Source: An Uncertainty Quantification Approach (TPAMI 2022 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Boosting Domain Adaptation by Discovering Latent Domains (CVPR 2018 · similarity 0.88)
- A Broad Study of Pre-training for Domain Generalization and Adaptation (ECCV 2022 · similarity 0.87)
- Information-Theoretic Regularization for Multi-Source Domain Adaptation (ICCV 2021 · similarity 0.87)
- A General Representation-Based Approach to Multi-Source Domain Adaptation (ICML 2025 · similarity 0.86)
- Co-regularized Alignment for Unsupervised Domain Adaptation (NeurIPS 2018 · similarity 0.86)
- SF(DA)$^2$: Source-free Domain Adaptation Through the Lens of Data Augmentation (ICLR 2024 · similarity 0.85)
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.
