NIPS 2017 — notes and thoughts 个人笔记

只写一下自己觉得很有趣的部分

keytrends 今年流行啥？

1. 深度学习还是很火。视觉图像等有了很多处理应用。多数是CNN的变种，比较牛逼的有Capsule Networks 和 WaveNet。
2. 强化学习
3. Meta-Learning and One-Shot learning
4. GANS
5. Bayesian NNs are area of active research
6. Fairness in ML
7. Explainable ML
8. 加速SGD的tricks
9. 图学习

关于机器学习的可解释性的思考？试图了解ML

有趣的项目/文章

• 强化学习买机票
• ML in credit, education, employment, housing, marketing verticals 机器学习用于信用，教育，就业，住宅，市场领域
• Powering the next 100 years/John Platt 展望未来100年
• Reprogramming Human Genome 对人类基因重新编程/编辑
• The Trouble with Bias 关于Biases的一些思考和讨论
• A Unified Approach To Interpreting Model Predictions 解释模型预测机制的一种统一方法，用MINIST，有git源码，非常直观
• Towards Accurate Binary Convolutional Neural Network (ABC-Net)
  Authors claim that this is the first time a binary neural network achieves prediction accuracy
  comparable to its full-precision counterpart on ImageNet.
• The Unreasonable Effectiveness of Structure 结构的不可思议的牛逼之处
• Engineering and Reverse-Engineering Intelligence Using Probabilistic Programs, Program Induction, and Deep Learning
• Attention is all you need 关于机器翻译
• A simple neural network module for relational reasoning 问答
• Train longer, generalize better: closing the generalization gap in large batch training of neural networks 加速SGD
• Deep Learning for Robotics 深度学习用在机器人上
• Learning State Representations
  “Shallow learning, deep representations” – sophisticated learning of problem representation makes learning task simple
  浅层学习，但是深度表示。问题表示很复杂，学起来很容易？
• AlphaZero – mastering games without human knowledge
  压轴戏，AlphaZero不需要人类知识就掌握了游戏？这篇文章详细地介绍了Zero是怎么做到的。

原文在这里：
https://olgalitech.wordpress.com/2017/12/12/nips-2017-notes-and-thoughs/