0429 SMT总结 Curriculum learning for MT

时间 2021-01-15

标签论文笔记繁體版

原文原文链接

课程学习（Curriculum Learning）由Montreal大学的Bengio教授团队在2009年的ICML上提出，主要思想是模仿人类学习的特点，由简单到困难来学习课程（在机器学习里就是容易学习的样本和不容易学习的样本），这样容易使模型找到更好的局部最优，同时加快训练的速度。如何在将Curriculum learning用于MT？总结与思考：（1）首先可以利用Curriculum l

>>阅读原文<<

1. （DCL）Dynamic Curriculum Learning for Imbalanced Data Classification
2. Segmentation Guided Attention Network for Crowd Counting via Curriculum Learning
3. 读文献——《Curriculum learning》
4. Learning Monocular Visual Odometrythrough Geometry-Aware Curriculum Learning
5. CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognition - 论文学习
6. 【论文解读 WSDM 2018 | DRL】Curriculum Learning for Heterogeneous Star Network Embedding via DRL
7. 【论文读后感】《Self-Attention Enhanced CNNs and Collaborative Curriculum Learning for Distantly Supervised》
8. Deep Residual Learning for Image Recognition总结
9. 《Wide & Deep Learning for Recommender Systems》论文总结
10. 【MT】牛津的MT教程
更多相关文章...
• Docker 资源汇总 - Docker教程
• Scala for循环 - Scala教程
• 算法总结-回溯法
• 算法总结-双指针

最新文章

1. 【Java8新特性_尚硅谷】P1_P5
2. SpringSecurity 基础应用
3. SlowFast Networks for Video Recognition
4. 074-enable-right-click
5. WindowFocusListener窗体焦点监听器
6. DNS部署（二）DNS的解析（正向、反向、双向、邮件解析及域名转换）
7. Java基础（十九）集合（1）集合中主要接口和实现类
8. 浏览器工作原理学习笔记
9. chrome浏览器构架学习笔记
10. eclipse引用sun.misc开头的类

本站公众号

欢迎关注本站公众号,获取更多信息

1. （DCL）Dynamic Curriculum Learning for Imbalanced Data Classification
2. Segmentation Guided Attention Network for Crowd Counting via Curriculum Learning
3. 读文献——《Curriculum learning》
4. Learning Monocular Visual Odometrythrough Geometry-Aware Curriculum Learning
5. CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognition - 论文学习
6. 【论文解读 WSDM 2018 | DRL】Curriculum Learning for Heterogeneous Star Network Embedding via DRL
7. 【论文读后感】《Self-Attention Enhanced CNNs and Collaborative Curriculum Learning for Distantly Supervised》
8. Deep Residual Learning for Image Recognition总结
9. 《Wide & Deep Learning for Recommender Systems》论文总结
10. 【MT】牛津的MT教程

>>更多相关文章<<