Re-weighted adversarial adaptation network for unsupervised domain adaptation
摘要主要強調以下幾點： 1. 當domain discrepancy比較大的時候，文章中方法的性能比其餘方法要好的多（實驗部分的確如此） 2. re-weighted：在訓練discriminator D的時候，每一個source sample並不是同等重要，weight取決於label 3. 在匹配source sample和target sample的特徵分佈的時候，用的是基於EM 距離的OT
相關文章
- 1. Domain Adaptation 與 Generative Adversarial Network
- 2. Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
- 3. [cvpr2017]Deep Hashing Network for Unsupervised Domain Adaptation
- 4. Multi-Adversarial Domain Adaptation
- 5. Geodesic flow kernel for unsupervised domain adaptation
- 6. [cvpr2017]Unsupervised Pixel–Level Domain Adaptation with Generative Adversarial Networks
- 7. ADDA: Adversarial Discriminative Domain Adaptation
- 8. [cvpr2017]Adversarial Discriminative Domain Adaptation
- 9. Unsupervised Domain Adaptation by Backpropagation(2015)
- 10. 2019 AAAI Transferable Attention for Domain Adaptation
- 更多相關文章...
- • Swift for 循環 - Swift 教程
- • Scala for循環 - Scala教程
- • Flink 數據傳輸及反壓詳解
- • 算法總結-廣度優先算法
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
最新文章
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. Domain Adaptation 與 Generative Adversarial Network
- 2. Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
- 3. [cvpr2017]Deep Hashing Network for Unsupervised Domain Adaptation
- 4. Multi-Adversarial Domain Adaptation
- 5. Geodesic flow kernel for unsupervised domain adaptation
- 6. [cvpr2017]Unsupervised Pixel–Level Domain Adaptation with Generative Adversarial Networks
- 7. ADDA: Adversarial Discriminative Domain Adaptation
- 8. [cvpr2017]Adversarial Discriminative Domain Adaptation
- 9. Unsupervised Domain Adaptation by Backpropagation(2015)
- 10. 2019 AAAI Transferable Attention for Domain Adaptation