年度回顾：2018年度机器学习50大热门网文-CSDN博客

云栖君导读：本文总结了2018年期间机器学习博客top50篇，在这个寒冬中给大家带来一点干粮。

新的一年新气象，总结过去一年，展望新的一年。站在巨人的肩膀上前行，肯定会事半功倍。因此，本文从2018年1月至12月期间挑选出近22,000篇机器学习文章，并进行比较，以挑选出能够提升2019年数据科学技能的前50名文章。

从概率上讲，这是一个极具竞争力的列表，概率仅为50 / 22,000（0.23％），且需要经过仔细挑选并与过去一年发布的机器学习文章进行对比。Mybridge AI通过考虑受欢迎程度、参与度和新近度以及其他人为因素来评估这些文章的质量。

本教程将50篇文章划分为16个相关组：

寒冬已至，请花些大量时间阅读过去一年中可能错过的顶级机器学习教程。如果想查看去年最好的机器学习系列文章，请点击这里。

https://medium.mybridge.co/learn-to-build-a-machine-learning-application-from-top-articles-of-2017-cdd5638453fc

深度视频

No. 1

Deepfakes与家庭乐趣，如何让自己妻子参加今夜秀——由Sven Charleer提供；

https://towardsdatascience.com/family-fun-with-deepfakes-or-how-i-got-my-wife-onto-the-tonight-show-a4454775c011?spm=a2c4e.11153940.blogcont683655.20.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 2

深度视频肖像：一种新颖的方法，只使用输入视频即可实现肖像视频的照片般逼真的重新动画——由Christian Theobalt提供；

https://web.stanford.edu/~zollhoef/papers/SG2018_DeepVideo/page.html%0A?spm=a2c4e.11153940.blogcont683655.21.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

人脸识别

No. 3

如何使用Python中的深度学习实现iPhone X的FaceID功能——Nouman Di PaloCourtesy；

https://towardsdatascience.com/how-i-implemented-iphone-xs-faceid-using-deep-learning-in-python-d5dbaa128e1d?spm=a2c4e.11153940.blogcont683655.22.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 4

使用OpenCV、Python和深度学习进行人脸识别——由Adrian Rosebrock提供；

https://www.pyimagesearch.com/2018/06/18/face-recognition-with-opencv-python-and-deep-learning?spm=a2c4e.11153940.blogcont683655.23.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 5

前沿人脸识别很复杂，这些电子表格让它变得更容易——由 Dave Smith提供；

https://towardsdatascience.com/cutting-edge-face-recognition-is-complicated-these-spreadsheets-make-it-easier-e7864dbf0e1a?spm=a2c4e.11153940.blogcont683655.24.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

对象检测

No. 6

在Airbnb上分类列表照片：大规模深度学习模型正在改变我们在平台上思考家庭图像的方式——由Shijing Yao提供；

https://medium.com/airbnb-engineering/categorizing-listing-photos-at-airbnb-f9483f3ab7e3?spm=a2c4e.11153940.blogcont683655.25.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 7

使用OpenCV进行YOLO对象检测——由Adrian Rosebrock提供；

https://www.pyimagesearch.com/2018/11/12/yolo-object-detection-with-opencv?spm=a2c4e.11153940.blogcont683655.26.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 8

使用10行代码实现对象检测——由Moses Olafenwa提供；

https://towardsdatascience.com/object-detection-with-10-lines-of-code-d6cb4d86f606?spm=a2c4e.11153940.blogcont683655.27.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

游戏AI

No. 9

游戏AI的初学者指南——由Kylotan提供；

https://www.gamedev.net/articles/programming/artificial-intelligence/the-total-beginners-guide-to-game-ai-r4942?spm=a2c4e.11153940.blogcont683655.28.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 10

基于预测奖励的强化学习——由Harri Edwards提供；

https://blog.openai.com/reinforcement-learning-with-prediction-based-rewards?spm=a2c4e.11153940.blogcont683655.29.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 11

Montezuma的复仇之路通过Go-Explore解决，这是一种新的解决困难探索问题的算法——由优步工程师提供；

https://eng.uber.com/go-explore?spm=a2c4e.11153940.blogcont683655.30.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 12

抢旗：代理如何在复杂的第一人称多人游戏中实现人类级别的表现，甚至可以与人类队友合作——由DeepMind提供；

https://deepmind.com/blog/capture-the-flag?spm=a2c4e.11153940.blogcont683655.31.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 13

OpenAI Five：在Dota 2游戏中击败业余人类玩家——由OpenAI提供；

https://blog.openai.com/openai-five?spm=a2c4e.11153940.blogcont683655.32.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

象棋

No. 14

AlphaZero：在国际象棋、将棋和围棋的盛大游戏中崭露头角——由DeepMind提供；

https://deepmind.com/blog/alphazero-shedding-new-light-grand-games-chess-shogi-and-go?spm=a2c4e.11153940.blogcont683655.33.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 15

如何使用Python和Keras构建自己的AlphaZero AI——由David Foster提供；

https://medium.com/applied-data-science/how-to-build-your-own-alphazero-ai-using-python-and-keras-7f664945c188?spm=a2c4e.11153940.blogcont683655.34.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 16

简单解释：人工智能程序如何掌握围棋游戏的古老游戏——由Aman Agarwal提供；

https://medium.freecodecamp.org/explained-simply-how-an-ai-program-mastered-the-ancient-game-of-go-62b8940a9080?spm=a2c4e.11153940.blogcont683655.35.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

医疗

No. 17

深度学习在医学图像数据集中的不合理用处——由Luke Oakden-Rayner提供；

https://lukeoakdenrayner.wordpress.com/2018/04/30/the-unreasonable-usefulness-of-deep-learning-in-medical-image-datasets?spm=a2c4e.11153940.blogcont683655.36.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 18

利用基于DNA的胜者通吃神经网络扩大分子模式识别——由Kevin M. Cherry、Lulu Qian提供；

https://www.nature.com/articles/s41586-018-0289-6.epdf?spm=a2c4e.11153940.blogcont683655.37.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 19

针对脑核磁共振图像的深度学习方法——由Henrik Marklund提供；

https://medium.com/stanford-ai-for-healthcare/its-a-no-brainer-deep-learning-for-brain-mr-images-f60116397472?spm=a2c4e.11153940.blogcont683655.38.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

运动

No. 20

每个人都跳舞：一个简单的方法“跟我做”动作迁移——由Caroline Chan等人提供；

https://carolineec.github.io/everybody_dance_now?spm=a2c4e.11153940.blogcont683655.39.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 21

走向虚拟替身演员——由Xue Bin peng提供；

https://bair.berkeley.edu/blog/2018/04/10/virtual-stuntman?spm=a2c4e.11153940.blogcont683655.40.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 22

学习敏捷：一个真正的机器人手，使用与OpenAI Five相同的学习算法和代码进行训练，已经学习了类似于旋转物体的动作——由OpenAI提供；

https://blog.openai.com/learning-dexterity?spm=a2c4e.11153940.blogcont683655.41.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 23

在人工代理中使用类似网格的表示进行导航——由Andrea Banino等人提供；

https://deepmind.com/blog/grid-cells?spm=a2c4e.11153940.blogcont683655.42.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

Web＆App

No. 24

如何使用CoreML、PyTorch和React Native在iOS上发布神经网络——由 Stefano等人提供；

https://attardi.org/pytorch-and-coreml?spm=a2c4e.11153940.blogcont683655.43.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 25

如何训练AI将设计模型转换为HTML和CSS——由Emil Wallner提供；

https://medium.freecodecamp.org/how-you-can-train-an-ai-to-convert-your-design-mockups-into-html-and-css-cc7afd82fed4?spm=a2c4e.11153940.blogcont683655.44.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

翻译

No. 26

通过更快的训练和推理将神经机器翻译成更大的数据集——由Michael Auli等人提供；

https://code.fb.com/ai-research/scaling-neural-machine-translation-to-bigger-data-sets-with-faster-training-and-inference?spm=a2c4e.11153940.blogcont683655.45.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 27

在翻译中找到：通过深入学习从头开始构建语言翻译——由Samuel Lynn-Evans等人提供；

https://blog.floydhub.com/language-translator?spm=a2c4e.11153940.blogcont683655.46.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 28

无监督机器翻译：为更多语言提供快速，准确翻译的新方法。由Facebook Research提供

https://code.fb.com/ai-research/unsupervised-machine-translation-a-novel-approach-to-provide-fast-accurate-translations-for-more-languages?spm=a2c4e.11153940.blogcont683655.47.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

NLP

No. 29

有关BERT、ELMo和co（如何NLP破解转移学习）的说明——由Jay Alammar提供

http://jalammar.github.io/illustrated-bert?spm=a2c4e.11153940.blogcont683655.48.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 30

注释迁移学习——由哈佛NLP组提供；

http://nlp.seas.harvard.edu/2018/04/03/attention.html?spm=a2c4e.11153940.blogcont683655.49.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 31

自然语言处理很有趣——由Adam Geitgey提供；

https://medium.com/@ageitgey/natural-language-processing-is-fun-9a0bff37854e?spm=a2c4e.11153940.blogcont683655.50.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

神经网络

No. 32

如何在Python中从头开始构建自己的神经网络——由James Loy提供；

https://towardsdatascience.com/how-to-build-your-own-neural-network-from-scratch-in-python-68998a08e4f6?spm=a2c4e.11153940.blogcont683655.51.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 33

使用简单的NumPy编写一个神经网络——由Piotr Skalski提供；

https://towardsdatascience.com/lets-code-a-neural-network-in-plain-numpy-ae7e74410795?spm=a2c4e.11153940.blogcont683655.52.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

CNN

No. 34

可区分的图像参数化：一种功能强大、探索不足的神经网络可视化和艺术工具——由distillpub提供；

https://distill.pub/2018/differentiable-parameterizations?spm=a2c4e.11153940.blogcont683655.53.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 35

特征转换——由distillpub提供；

https://distill.pub/2018/feature-wise-transformations?spm=a2c4e.11153940.blogcont683655.54.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 36

Keras和卷积神经网络（CNN——由Adrian Rosebrock提供；

https://www.pyimagesearch.com/2018/04/16/keras-and-convolutional-neural-networks-cnns?spm=a2c4e.11153940.blogcont683655.55.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 37

可解释性的组成部分——由Distill提供；

https://distill.pub/2018/building-blocks?spm=a2c4e.11153940.blogcont683655.56.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 38

Rosetta：通过机器学习理解图像和视频中的文本——由Facebook Research提供；

https://code.fb.com/ai-research/rosetta-understanding-text-in-images-and-videos-with-machine-learning?spm=a2c4e.11153940.blogcont683655.57.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 39

一个有趣的卷积神经网络失败案例和协同解决方案——由优步提供；

https://eng.uber.com/coordconv?spm=a2c4e.11153940.blogcont683655.58.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

RNN

No. 40

Google Duplex：用于通过电话完成真实世界任务的AI系统——由Yaniv Leviathan提供；

https://ai.googleblog.com/2018/05/duplex-ai-system-for-natural-conversation.html?spm=a2c4e.11153940.blogcont683655.59.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 41

世界模型：代理人可以在自己的梦中学习吗？——由maru提供；

https://worldmodels.github.io/?spm=a2c4e.11153940.blogcont683655.60.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

强化学习

No. 42

经验教训再现深度强化学习论文——由Matthew Rahtz提供；

http://amid.fish/reproducing-deep-rl?spm=a2c4e.11153940.blogcont683655.61.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 43

具有强化学习的灵巧操作：高效、通用和低成本——由Henry Zhu等人提供；

https://bair.berkeley.edu/blog/2018/08/31/dexterous-manip?spm=a2c4e.11153940.blogcont683655.62.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 44

深度强化学习不起作用——由Sorta Insightful
提供；

https://www.alexirpan.com/2018/02/14/rl-hard.html?spm=a2c4e.11153940.blogcont683655.63.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

TensorFlow

No. 45

TensorFlow中的三元组损失和在线挖掘——由Olivier Moindrot提供；

https://omoindrot.github.io/triplet-loss?spm=a2c4e.11153940.blogcont683655.64.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 46

Tensorflow：令人困惑的部件（1）——由Jacob Buckman提供；

http://jacobbuckman.com/post/tensorflow-the-confusing-parts-1?spm=a2c4e.11153940.blogcont683655.65.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 47

Tensorflow-Project-Template：TensorFlow项目模板架构的最佳实践（Github上已有2579颗星）——由Mahmoud Gemy提供；

https://github.com/Mrgemy95/Tensorflow-Project-Template?spm=a2c4e.11153940.blogcont683655.66.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 48

使用TensorFlow.js在浏览器中进行实时人体姿态估计——由TensorFlow提供；

https://medium.com/tensorflow/real-time-human-pose-estimation-in-the-browser-with-tensorflow-js-7dd0bc881cd5?spm=a2c4e.11153940.blogcont683655.67.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

指南

No. 49

机器学习规则：| ML通用指南|谷歌开发者——由Martin Zinkevich提供；

https://developers.google.com/machine-learning/guides/rules-of-ml?spm=a2c4e.11153940.blogcont683655.68.1e195250rLB3BI&utm_source=mybridge&utm_medium=blog&utm_campaign=read_more

No. 50

基于模型的机器学习——由John Winn和Christopher M. Bishop提供