Research map / Label-efficient and robust learning / Domain adaptation and generalization
Residual Parameter Transfer for Deep Domain Adaptation
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 |
|---|---|---|---|
| use residual parameter transfer with automatic complexity reduction instead of fixed-complexity transfer (ablation) | Ours on SVHN→MNIST with SvhNet · SVHN→MNIST | accuracy: 77.8 [0.09] -> 78.7 [0.12] | Helped |
| use residual parameter transfer rather than domain-invariant DANN features | SVHN→MNIST with SvhNet · SVHN→MNIST | accuracy: 73.9 [0.79] -> 78.7 [0.12] | Helped |
| use residual parameter transfer rather than the strongest listed baseline, Two-stream | SVHN→MNIST with LeNeT · SVHN→MNIST | accuracy: 82.8 [0.20] -> 84.7 [0.17] | Helped |
| use residual parameter transfer rather than two-stream adaptation | Ours on UAV-200 · UAV-200 | Average Precision: .732 [.003] -> .743 [.006] | Helped |
| use residual parameter transfer rather than DANN | Ours on Office A→D · Office A→D | accuracy: 79.1 -> 82.7 [0.3] | Helped |
| use residual parameter transfer rather than DANN | Ours on Office D→A · Office D→A | accuracy: 63.6 -> 64.7 [0.2] | Helped |
| use residual parameter transfer rather than DANN | Ours on Office A→W · Office A→W | accuracy: 78.9 -> 81.5 [0.7] | Helped |
| use residual parameter transfer rather than DANN | Ours on Office W→A · Office W→A | accuracy: 62.8 -> 63.6 [0.2] | Helped |
| use residual parameter transfer rather than DANN | Ours on Office D→W · Office D→W | accuracy: 97.5 -> 98.0 [0.1] | Helped |
| use residual parameter transfer rather than DANN | Ours on Office W→D · Office W→D | accuracy: 99.2 -> 99.4 [0.1] | Helped |
| use residual parameter transfer with an RMAN-based discrepancy loss instead of RMAN with shared parameters | SVHN→MNIST with SvhNet · SVHN→MNIST | accuracy: 81.0 [0.77] -> 84.6 [1.26] | Helped |
| set the stream regularization weight λ_s to zero (ablation) | residual parameter transfer · domain adaptation | accuracy | Hurt |
Closest papers by meaning
- Unsupervised Domain Adaptation with Residual Transfer Networks (NeurIPS 2016 · similarity 0.93)
- Interpretations of Domain Adaptations via Layer Variational Analysis (ICLR 2023 · similarity 0.90)
- Deep Asymmetric Transfer Network for Unbalanced Domain Adaptation (AAAI 2018 · similarity 0.89)
- Reducing Divergence in Batch Normalization for Domain Adaptation (AAAI 2025 · similarity 0.87)
- A General Representation-Based Approach to Multi-Source Domain Adaptation (ICML 2025 · similarity 0.87)
- Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit (ICLR 2024 · similarity 0.87)
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.