Be Your Own Teacher: Improve the Performance of Convolutional Neural Networks via Self Distillation
Loss Source 1: Cross entropy loss，各個階段的分類器都有 Loss Source 2: KL loss，深層的分類器作爲淺層分類器的teacher Loss Source 3: L2 loss from hints，深層分類器的特徵和淺層分類器的特徵做L2 loss，bottleneck即feature adaptation，爲了使student和teacher一樣
相關文章
- 1. Awesome Knowledge-Distillation
- 2. Bag of Tricks for Convolutional Neural Networks
- 3. Convolutional Neural Networks---Foundations of Convolutional Neural Networks
- 4. Convolutional Neural Networks: Application
- 5. The Unreasonable Effectiveness of Recurrent Neural Networks
- 6. Understanding Convolutional Neural Networks for NLP
- 7. [轉] Understanding Convolutional Neural Networks for NLP
- 8. Using AJAX to Improve the Bandwidth Performance of Web Applications
- 9. [C6] Andrew Ng - Convolutional Neural Networks
- 10. Bag of Tricks for Image Classification with Convolutional Neural Networks
- 更多相關文章...
- • XSLT
- • XSLT
- • 爲了進字節跳動，我精選了29道Java經典算法題，帶詳細講解
- • Java 8 Stream 教程
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. Awesome Knowledge-Distillation
- 2. Bag of Tricks for Convolutional Neural Networks
- 3. Convolutional Neural Networks---Foundations of Convolutional Neural Networks
- 4. Convolutional Neural Networks: Application
- 5. The Unreasonable Effectiveness of Recurrent Neural Networks
- 6. Understanding Convolutional Neural Networks for NLP
- 7. [轉] Understanding Convolutional Neural Networks for NLP
- 8. Using AJAX to Improve the Bandwidth Performance of Web Applications
- 9. [C6] Andrew Ng - Convolutional Neural Networks
- 10. Bag of Tricks for Image Classification with Convolutional Neural Networks