参考资料

1.Machine Learning[book]

Tom M.Mitchell,Chapter 1-1.1,1997

2.Tidy Data[paper]

Hadley Wickham,Journal of Statistical Software,August 2014,Volume 59,Issue 10

3.Prediction of Sea Surface Temperature using Long Short-Term Memory[paper]

Qin Zhang,Hui Wang,Junyu Dong,Guoqiang Zhong,Xin Su,19 May 2017,arXiv:1705.06861

4.ImageNet Classification with Deep Convolutional Neural Networks[paper]

Alex Krizhevsky,Ilya Sutskever,Geoffrey E.Hinton,NISP 2012

5.Deep Unsupervised Clustering Using Mixture of Autoencoders[paper]

Dejiao Zhang,Yifan Sun,Brian Eriksson,Laura Balzano.21 Dec 2017,arXiv:1712.07788

6.Good Semi-supervised Learning that Requires a Bad GAN[paper]

Zihang Dai,Ruslan Salakhutdinov,et.al.3 Nov 2017,arXiv:1705.09783

7.Playing Atari with Deep Reinforcement Learning[paper]

Volodymyr Mnih,et.al.19 Dec 2013.arXiv:1312.5602

8.《迁移学习简明手册——一点心得体会》V1.0[notes]

王晋东,中国科学院计算技术研究所,2018.4

9.A Survey on Transfer Learning[paper]

Sinno Jialin Pan,Qiang Yang,IEEE 2009


[1]MECE分析法是麦肯锡思维的准则,即Mutually Exclusive Collective Exhaustive,意思是相互独立,完全穷尽。

[2]RGB色彩模式是工业界的一种颜色标准,是通过对红色(R)、绿色(G)、蓝色(B)三种颜色通道的变换以及它们之间的叠加来得到各种颜色。

[3]“三双”指的是得分、篮板次数和助攻次数都大于10。

[4]查准率和查全率也被称为准确率和召回率,但笔者觉得前者的定义更直观,因为查准率就是查准的(准=查出的内容有多少是有用的),而查全率就是查全的(全=有用的内容有多少被查出)。