机器学习人工学weekly-2018/9/2

注意下面很多链接需要访问外国网站，无奈国情如此

1. Google开源RL快速原型框架Dopamine

Introducing a New Framework for Flexible and Reproducible Reinforcement Learning Research

链接：

https://ai.googleblog.com/2018/08/introducing-new-framework-for-flexible.html

2. Google Assistant支持同时语言混搭对话

Teaching the Google Assistant to be Multilingual

链接：

https://ai.googleblog.com/2018/08/Multilingual-Google-Assistant.html

3. Google TPU动画介绍

What makes TPUs fine-tuned for deep learning?

链接：

https://cloud.google.com/blog/products/ai-machine-learning/what-makes-tpus-fine-tuned-for-deep-learning

动画展示：

https://tpudemo.com/

4. Uber用深度学习替代一般机器学习做客服ticket

Scaling Uber’s Customer Support Ticket Assistant (COTA) System with Deep Learning

链接：

https://eng.uber.com/cota-v2/

5. BAIR文章，用RL快速学习机械手动作

Dexterous Manipulation with Reinforcement Learning: Efficient, General, and Low-Cost

链接：

https://bair.berkeley.edu/blog/2018/08/31/dexterous-manip/

6. federated learning介绍

链接：

https://florian.github.io/federated-learning/

7. 3D物体

Beyond the pixel plane: sensing and learning in 3D

链接：

https://thegradient.pub/beyond-the-pixel-plane-sensing-and-learning-in-3d/

8. NLP泛化问题

NLP’s generalization problem, and how researchers are tackling it

链接：

https://thegradient.pub/frontiers-of-generalization-in-natural-language-processing/

9. 重现论文实现，RL learn2learn经典算法MAML和Reptile

Paper repro: Deep Metalearning using “MAML” and “Reptile”

链接：

https://towardsdatascience.com/paper-repro-deep-metalearning-using-maml-and-reptile-fd1df1cc81b0