Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learning
arxiv鏈接 前言 有時候, 輸入數據加入細微的擾動, 都會讓模型結果大相徑庭. 因此, 我們總是希望訓練結果的分佈是光滑的, 這樣才能保持穩定. 但在連續分佈的空間中, 總是存在過大的間隙, 這些間隙只能靠更多的數據或減小網絡複雜度(正則)的方法讓它儘可能光滑. VAT同樣是爲了解決光滑問題, 且解釋性更強, 也更直觀. 我只讀了文章前4頁, 做個大概瞭解. 思路 一個圖片是N維中的一個點P,
相關文章
- 1. Regularization method for machine learning
- 2. Adversarial learning for semi-supervised semantic segmentation
- 3. DropBlock： A regularization method for convolutional networks
- 4. 5.Adversarial Training Methods for Semi-Supervised Text Classification
- 5. Pseudo-Label：The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks
- 6. 論文共讀 | 「陽奉陰違」的半監督學習算法 - Virtual Adversarial Training
- 7. 《Adversarial learning for weakly-supervised social network alignment》翻譯
- 8. 論文：Adversarial Learning for Semi-Supervised Semantic Segmentation
- 9. 《DropBlock: A regularization method for convolutional networks》筆記
- 10. （原）DropBlock A regularization method for convolutional networks
- 更多相關文章...
- • Scala for循環 - Scala教程
- • Lua for 循環 - Lua 教程
- • RxJava操作符（七）Conditional and Boolean
- • Java Agent入門實戰（一）-Instrumentation介紹與使用
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。
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. Regularization method for machine learning
- 2. Adversarial learning for semi-supervised semantic segmentation
- 3. DropBlock： A regularization method for convolutional networks
- 4. 5.Adversarial Training Methods for Semi-Supervised Text Classification
- 5. Pseudo-Label：The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks
- 6. 論文共讀 | 「陽奉陰違」的半監督學習算法 - Virtual Adversarial Training
- 7. 《Adversarial learning for weakly-supervised social network alignment》翻譯
- 8. 論文：Adversarial Learning for Semi-Supervised Semantic Segmentation
- 9. 《DropBlock: A regularization method for convolutional networks》筆記
- 10. （原）DropBlock A regularization method for convolutional networks