Neural Headline Generation with Minimum Risk Training
今天分享的paper是Neural Headline Generation with Minimum Risk Training。 本文通過將評價指標融入目標函數來訓練模型，在中文和英文數據集上均取得了超過之前所有模型的結果。結果一點也不意外，因爲傳統的MLE並不是以ROUGE評價指標最大爲目標函數，而本文的方法針對了評價指標來做文章，一定會得到不錯的結果。反過來，我們需要思考一個問題，如果文本摘
相關文章
- 1. 【讀】seq2seq——（6）Neural Headline Generation with Minimum Risk Training
- 2. Text Generation With LSTM Recurrent Neural Networks in Python with Keras
- 3. Neural Response Generation via GAN with an Approximate Embedding Layer
- 4. BinaryConnect： Training Deep Neural Networks with binary weights during propagations
- 5. TRAINING DEEP NEURAL NETWORKS WITH LOW PRECISION MULTIPLICATIONS
- 6. Synchronous Bidirectional Inference for Neural Sequence Generation
- 7. Training Neural Networks, part I
- 8. Training Recurrent Neural Network
- 9. （轉）A Recipe for Training Neural Networks
- 10. AAAI2018-Long Text Generation via Adversarial Training with Leaked Information論文筆記
- 更多相關文章...
- • XSLT
- • Docker 容器連接 - Docker教程
- • 爲了進字節跳動，我精選了29道Java經典算法題，帶詳細講解
- • Composer 安裝與使用
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。
最新文章
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. 【讀】seq2seq——（6）Neural Headline Generation with Minimum Risk Training
- 2. Text Generation With LSTM Recurrent Neural Networks in Python with Keras
- 3. Neural Response Generation via GAN with an Approximate Embedding Layer
- 4. BinaryConnect： Training Deep Neural Networks with binary weights during propagations
- 5. TRAINING DEEP NEURAL NETWORKS WITH LOW PRECISION MULTIPLICATIONS
- 6. Synchronous Bidirectional Inference for Neural Sequence Generation
- 7. Training Neural Networks, part I
- 8. Training Recurrent Neural Network
- 9. （轉）A Recipe for Training Neural Networks
- 10. AAAI2018-Long Text Generation via Adversarial Training with Leaked Information論文筆記