Neural Response Generation via GAN with an Approximate Embedding Layer
1. Goal 用GAN來對dialogue system建模，從而減少safe response（e.g. 「我不知道」, 「我也是」）。 2. Contribution 提出了一個AEL (Approximate Embedding Layer )來解決seq2seq模型decode階段由於sampling導致的不可微問題（non-differentiable problem）。 3. Mod
相關文章
- 1. 系列總結：Sequence Generation via GAN
- 2. Neural Headline Generation with Minimum Risk Training
- 3. Embedding Layer
- 4. Text Generation With LSTM Recurrent Neural Networks in Python with Keras
- 5. Fast Approximate Energy Minimization via Graph Cuts
- 6. (zhuan) Recurrent Neural Network
- 7. 論文閱讀：Neural Image Caption Generation with Visual Attention
- 8. Deep Speaker: an End-to-End Neural Speaker Embedding System
- 9. 【讀】seq2seq——（6）Neural Headline Generation with Minimum Risk Training
- 10. GAN(9)——sequence generation
- 更多相關文章...
- • ASP Response 對象 - ASP 教程
- • XSLT
- • Flink 數據傳輸及反壓詳解
- • 爲了進字節跳動，我精選了29道Java經典算法題，帶詳細講解
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. 系列總結：Sequence Generation via GAN
- 2. Neural Headline Generation with Minimum Risk Training
- 3. Embedding Layer
- 4. Text Generation With LSTM Recurrent Neural Networks in Python with Keras
- 5. Fast Approximate Energy Minimization via Graph Cuts
- 6. (zhuan) Recurrent Neural Network
- 7. 論文閱讀：Neural Image Caption Generation with Visual Attention
- 8. Deep Speaker: an End-to-End Neural Speaker Embedding System
- 9. 【讀】seq2seq——（6）Neural Headline Generation with Minimum Risk Training
- 10. GAN(9)——sequence generation