Group-wise Contrastive Learning for Neural Dialogue Generation 阅读笔记
摘要 为解决生成多样性低的问题 思想 找到正向和负向group的context和response，设计loss 做法 首先有 其中pn是baseline模型，pm是target模型 其中c是输入文本 r是回复文本， 构造loss： 最大化正向回复，最小化负向回复 这个loss即刻画 相对baseline模型的生成多样性 或 分布distinctiveness 正向回复和负向回复的采样，使用了BM2
相关文章
- 1. Deep Reinforcement Learning for Dialogue Generation阅读笔记
- 2. Adversarial Learning for Neural Dialogue Generation 代码分析
- 3. Deep Reinforcement Learning for Dialogue Generation
- 4. Deep Reinforcement Learning for Dialogue Generation 论文阅读 A Diversity-Promoting Objective Function fo
- 5. 论文分享-- >Adversarial Learning for Neural Dialogue Generation
- 6. 论文阅读笔记《Meta-Learning of Neural Architectures for Few-Shot Learning》
- 7. 阅读《Learning to Ask: Neural Question Generation for Reading Comprehension 》
- 8. 论文阅读笔记《Attentive Weights Generation for Few Shot Learning via Information Maximization》
- 9. 《Contextualized Code Representation Learning for Commit Message Generation》阅读
- 10. Matching Networks for one Shot Learning 阅读笔记
- 更多相关文章...
- • RSS 阅读器 - RSS 教程
- • PHP 实例 - AJAX RSS 阅读器 - PHP教程
- • Tomcat学习笔记（史上最全tomcat学习笔记）
- • JDK13 GA发布：5大特性解读
相关标签/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。
欢迎关注本站公众号,获取更多信息
相关文章
- 1. Deep Reinforcement Learning for Dialogue Generation阅读笔记
- 2. Adversarial Learning for Neural Dialogue Generation 代码分析
- 3. Deep Reinforcement Learning for Dialogue Generation
- 4. Deep Reinforcement Learning for Dialogue Generation 论文阅读 A Diversity-Promoting Objective Function fo
- 5. 论文分享-- >Adversarial Learning for Neural Dialogue Generation
- 6. 论文阅读笔记《Meta-Learning of Neural Architectures for Few-Shot Learning》
- 7. 阅读《Learning to Ask: Neural Question Generation for Reading Comprehension 》
- 8. 论文阅读笔记《Attentive Weights Generation for Few Shot Learning via Information Maximization》
- 9. 《Contextualized Code Representation Learning for Commit Message Generation》阅读
- 10. Matching Networks for one Shot Learning 阅读笔记