图0:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习

整理 | 胡永波

根据《纽约时报》的说法,“在硅谷招募机器学习工程师、数据科学家的情形,越来越像NFL选拔职业运动员,没有苛刻的训练很难上场了。”毕竟,高达124472美元的平均年薪可不是谁想挣就能挣到的。

正如职业运动员每天都要训练一样,机器学习的日常练习也是工程师生涯得以大踏步前进的基本保障。仅2017年一年,机器学习领域总结此类实战经验的文章便已超过20000篇,该领域相关职位的热度自是可见一斑。

从中,我们筛选出50篇最好的经验和心得,囊括了机器学习在15大细分领域的各项典型应用:

图1:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习

  1. 图像处理
  2. 风格迁移
  3. 图像分类
  4. 面部识别
  5. 视频稳像
  6. 目标检测
  7. 自动驾驶
  8. 推荐系统
  9. AI游戏
  10. AI棋手
  11. AI医疗
  12. AI语音
  13. AI音乐
  14. 自然语言处理
  15. 学习预测

当然,如果你只是一个刚要准备上手机器学习的新人,我们推荐你优先考虑以下两个高分实战课程:

A) AI游戏【推荐:5041;评分:4.7/5】

图2:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习

The Beginner’s Guide to Building an Artificial Intelligence in Unity

  • 链接:http://www.udemy.com/artificial-intelligence-in-unity/

B) 计算机视觉【推荐:8161;评分:4.5/5】

图3:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习

Deep Learning and Computer Vision A-Z™: Learn OpenCV, SSD & GANs and create image recognition apps

  • 链接:http://www.udemy.com/computer-vision-a-z/

而对具体的实战经验,接下来我们分领域一一来看:

图像处理

1、High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs

  • GitHub:http://github.com/NVIDIA/pix2pixHD
  • 论文:http://arxiv.org/abs/1711.11585
  • 博客:http://tcwang0509.github.io/pix2pixHD/

图4:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:NVIDIA & UC Berkeley

2、Using Deep Learning to Create Professional-Level Photographs

  • GitHub:http://github.com/google/creatism
  • 论文:http://arxiv.org/abs/1707.03491
  • 博客:http://research.googleblog.com/2017/07/using-deep-learning-to-create.html

图5:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Google Research

3、High Dynamic Range (HDR) Imaging using OpenCV (Python)

  • 项目:http://www.learnopencv.com/high-dynamic-range-hdr-imaging-using-opencv-cpp-python/
  • 课程主页:http://courses.learnopencv.com/p/opencv-for-beginners

图6:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Satya Mallick

风格迁移

4、Visual Attribute Transfer through Deep Image Analogy

  • GitHub:http://github.com/msracver/Deep-Image-Analogy
  • 论文:http://arxiv.org/abs/1705.01088

图7:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:微软研究院 & 上海交大

5、Deep Photo Style Transfer

  • GitHub:http://github.com/luanfujun/deep-photo-styletransfer
  • 论文:http://arxiv.org/abs/1703.07511

图8:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Cornell University & Adobe

6、Deep Image Prior

  • GitHub:http://github.com/DmitryUlyanov/deep-image-prior
  • 论文:http://arxiv.org/abs/1711.10925
  • 博客:http://dmitryulyanov.github.io/deep_image_prior

图9:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:SkolTech & Yandex & Oxford University

图像分类

7、Feature Visualization: How neural networks build up their understanding of images.

  • 论文:http://distill.pub/2017/feature-visualization/
  • 代码:http://github.com/tensorflow/tensorflow/blob/master/tensorflow/examples/tutorials/deepdream/deepdream.ipynb
  • 博客:http://colah.github.io/

图10:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Google Brain

8、An absolute beginner’s guide to Image Classification with Neural Networks

  • Github【4491收藏】:http://github.com/humphd/have-fun-with-machine-learning
  • 中文版:http://github.com/humphd/have-fun-with-machine-learning/blob/master/README_zh-tw.md

图11:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Mozilla

9、Background removal with deep learning

  • 模型:http://towardsdatascience.com/background-removal-with-deep-learning-c4f2104b3157
  • 部署:http://medium.com/@burgalon/deploying-your-keras-model-35648f9dc5fb

图12:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Gidi Shperber

面部识别

10、Large Pose 3D Face Reconstruction from a Single Image via Direct Volumetric CNN Regression

  • GitHub:http://github.com/AaronJackson/vrn
  • 论文:http://arxiv.org/abs/1703.07834
  • 博客:http://aaronsplace.co.uk/papers/jackson2017recon/
  • Demo:http://cvl-demos.cs.nott.ac.uk/vrn/

图13:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Aaron Jackson

11、Eye blink detection with OpenCV, Python, and dlib

  • 项目:http://www.pyimagesearch.com/2017/04/24/eye-blink-detection-opencv-python-dlib/
  • 论文:http://vision.fe.uni-lj.si/cvww2016/proceedings/papers/05.pdf

图14:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习

作者:Adrian Rosebrock

12、DEAL WITH IT in Python with Face Detection

  • GitHub:http://github.com/burningion/automatic-memes
  • 博客:http://www.makeartwithpython.com/blog/deal-with-it-generator-face-recognition/

图15:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Kirk Kaiser

视频稳像

13、Fused Video Stabilization on the Pixel 2 and Pixel 2 XL

  • 博客:http://research.googleblog.com/2017/11/fused-video-stabilization-on-pixel-2.html
  • 测评:http://www.dxomark.com/google-pixel-2-reviewed-sets-new-record-smartphone-camera-quality/

图16:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Google Research

目标检测

14、How HBO’s Silicon Valley built “Not Hotdog” with mobile TensorFlow and Keras

  • 博客:http://medium.com/@timanglade/how-hbos-silicon-valley-built-not-hotdog-with-mobile-tensorflow-keras-react-native-ef03260747f3
  • 项目:http://github.com/kmather73/NotHotdog-Classifier

图17:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Tim Anglade

15、Object detection: an overview in the age of Deep Learning

  • GitHub:http://github.com/tryolabs/luminoth
  • 论文:http://tryolabs.com/blog/2017/08/30/object-detection-an-overview-in-the-age-of-deep-learning/

图18:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Tryolabs

16、How to train your own Object Detector with TensorFlow’s Object

Detector API

  • 博客:http://towardsdatascience.com/how-to-train-your-own-object-detector-with-tensorflows-object-detector-api-bec72ecfe1d9
  • 数据集:http://github.com/datitran/raccoon_dataset
  • 产品化:http://towardsdatascience.com/building-a-real-time-object-recognition-app-with-tensorflow-and-opencv-b7a2b4ebdc32
  • 产品代码:http://github.com/datitran/object_detector_app

图19:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Dat Tran

17、Real-time object detection with deep learning and OpenCV

  • 实战:http://www.pyimagesearch.com/2017/09/18/real-time-object-detection-with-deep-learning-and-opencv/
  • 入门:

①http://www.pyimagesearch.com/2017/09/11/object-detection-with-deep-learning-and-opencv/;②http://www.pyimagesearch.com/2016/01/04/unifying-picamera-and-cv2-videocapture-into-a-single-class-with-opencv/

③http://www.pyimagesearch.com/2017/08/21/deep-learning-with-opencv/

图20:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Adrian Rosebrock

自动驾驶

18、Self-driving Grand Theft Auto V with Python : Intro [Part I]

  • GitHub:http://github.com/sentdex/pygta5
  • 视频:http://www.youtube.com/playlist?list=PLQVvvaa0QuDeETZEOy4VdocT7TOjfSA8a
  • 博客:http://pythonprogramming.net/game-frames-open-cv-python-plays-gta-v/

图21:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Sentdex

19、Recognizing Traffic Lights With Deep Learning: How I learned deep learning in 10 weeks and won $5,000

  • GitHub:http://github.com/davidbrai/deep-learning-traffic-lights
  • 博客:http://medium.freecodecamp.org/recognizing-traffic-lights-with-deep-learning-23dae23287cc
  • 相关比赛:http://www.getnexar.com/challenge-1/

图22:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:David Brailovsky

推荐系统

20、Spotify’s Discover Weekly: How machine learning finds your new music

  • 实战:http://hackernoon.com/spotifys-discover-weekly-how-machine-learning-finds-your-new-music-19a41ab76efe
  • 演讲:http://www.youtube.com/watch?v=A259Yo8hBRs
  • 相关博客:②http://notes.variogr.am/2012/12/11/how-music-recommendation-works-and-doesnt-work/

图23:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Sophia Ciocca

21、Artwork Personalization at Netflix

  • 博客:http://medium.com/netflix-techblog/artwork-personalization-c589f074ad76
  • 论文:http://arxiv.org/abs/1003.5956
  • 原理介绍:http://highscalability.com/blog/2017/12/11/netflix-what-happens-when-you-press-play.html

图24:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Netflix

AI游戏

22、MariFlow — Self-Driving Mario Kart w/Recurrent Neural Network

  • 文档:http://docs.google.com/document/d/1p4ZOtziLmhf0jPbZTTaFxSKdYqE91dYcTNqTVdd6es4
  • 视频:http://www.youtube.com/watch?v=Ipi40cb_RsI

图25:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习

作者:SethBling

 

23、OpenAI Baselines: DQN

  • GitHub:http://github.com/openai/baselines
  • 项目主页:http://blog.openai.com/openai-baselines-dqn/

图26:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:OpenAI

24、Reinforcement Learning on Dota 2 [Part II]

  • 博客:http://blog.openai.com/more-on-dota-2/
  • 视频:http://openai.com/the-international/

图27:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:OpenAI

25、Creating an AI DOOM bot

  • 博客:http://www.codelitt.com/blog/doom-ai/
  • 工具:http://vizdoom.cs.put.edu.pl/

图28:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Abel Castilla

26、Phase-Functioned Neural Networks for Character Control

  • 博客:http://theorangeduck.com/page/phase-functioned-neural-networks-character-control
  • 代码:http://theorangeduck.com/media/uploads/other_stuff/pfnn.zip
  • 论文:http://theorangeduck.com/media/uploads/other_stuff/phasefunction.pdf
  • 视频:http://theorangeduck.com/media/uploads/other_stuff/phasefunction.mov

图29:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Daniel Holden

27、The Game Imitation: Deep Supervised Convolutional Networks for Quick Video Game AI

  • 论文:http://arxiv.org/abs/1702.05663
  • 视频:http://www.youtube.com/playlist?list=PLegUCwsQzmnUpPwVv8ygMa19zNnDgJ6OC

图30:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Stanford

28、Introducing: Unity Machine Learning Agents

  • GitHub:http://github.com/Unity-Technologies/ml-agents
  • 博客:http://blogs.unity3d.com/cn/2017/09/19/introducing-unity-machine-learning-agents/
  • 文档:http://github.com/Unity-Technologies/ml-agents/tree/master/docs

图31:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Unity

AI棋手

29、Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm

  • 论文:http://arxiv.org/abs/1712.01815
  • 演讲:http://ktiml.mff.cuni.cz/~bartak/ui_seminar/talks/2017ZS/KarelHa_AlphaZero.pdf模型:http://deepmind.com/research/alphago/alphazero-resources/
  • 相关实现:②http://web.stanford.edu/~surag/posts/alphazero.html

图32:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Deepmind

30、AlphaGo Zero: Learning from scratch

  • 博客:http://deepmind.com/blog/alphago-zero-learning-scratch/
  • 论文:http://deepmind.com/documents/119/agz_unformatted_nature.pdf
  • 棋谱:http://www.alphago-games.com/

图33:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:DeepMind

31、How Does DeepMind’s AlphaGo Zero Work?

  • GitHub:http://github.com/llSourcell/alphago_demo
  • 视频:http://www.youtube.com/watch?v=vC66XFoN4DE

图34:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Siraj Raval

32、A step-by-step guide to building a simple chess AI

  • GitHub:http://github.com/lhartikk/simple-chess-ai
  • 博客:http://medium.freecodecamp.org/simple-chess-ai-step-by-step-1d55a9266977
  • Wiki:http://chessprogramming.wikispaces.com/

图35:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Lauri Hartikka

AI医疗

33、CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning

  • 项目主页:http://stanfordmlgroup.github.io/projects/chexnet/
  • 论文:http://arxiv.org/abs/1711.05225
  • 博客:http://lukeoakdenrayner.wordpress.com/2017/11/18/quick-thoughts-on-chestxray14-performance-claims-and-clinical-tasks/

图36:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:吴恩达 & Stanford ML Group

34、Can you improve lung cancer detection? 2nd place solution for the Data Science Bowl 2017

  • Kaggle:http://www.kaggle.com/c/data-science-bowl-2017
  • GitHub:http://github.com/dhammack/DSB2017/
  • 博客:http://juliandewit.github.io/kaggle-ndsb2017/

图37:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Julian de Wit

35、Improving Palliative Care with Deep Learning

  • 项目主页:http://stanfordmlgroup.github.io/projects/improving-palliative-care/
  • 论文:http://arxiv.org/abs/1711.06402

图38:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:吴恩达 & Stanford ML Group

36、Heart Disease Diagnosis with Deep Learning

  • GitHub:http://github.com/chuckyee/cardiac-segmentation
  • 博客:http://blog.insightdatascience.com/heart-disease-diagnosis-with-deep-learning-c2d92c27e730
  • 文章:http://chuckyee.github.io/cardiac-segmentation/

图39:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Chuck-Hou Yee

AI语音

37、Tacotron: A Fully End-to-End Text-To-Speech Synthesis Model

  • GitHub:http://github.com/Kyubyong/tacotron
  • 论文:http://arxiv.org/abs/1703.10135
  • 项目主页:http://google.github.io/tacotron/

图40:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Google

38、Sequence Modeling with CTC

  • GitHub:http://github.com/awni/speech
  • 论文:http://distill.pub/2017/ctc/

图41:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Awni Hannun

39、Deep Voice: Real-time Neural Text-to-Speech

  • GitHub:http://github.com/israelg99/deepvoice
  • 论文:http://arxiv.org/abs/1702.07825
  • 博客:http://research.baidu.com/deep-voice-production-quality-text-speech-system-constructed-entirely-deep-neural-networks/

图42:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:百度

40、Deep Learning for Siri’s Voice: On-device Deep Mixture Density Networks for Hybrid Unit Selection Synthesis

  • 博客:http://machinelearning.apple.com/2017/08/06/siri-voices.html

图43:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习

来源:Apple

AI音乐

41、Computer evolves to generate baroque music!

  • 视频:http://www.youtube.com/watch?v=SacogDL_4JU
  • 相关博客:http://karpathy.github.io/2015/05/21/rnn-effectiveness/

图44:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Cary Huang

42、Make your own music with WaveNets: Making a Neural Synthesizer Instrument

  • GitHub:http://github.com/tensorflow/magenta/tree/master/magenta/models/nsynth
  • 论文:http://arxiv.org/abs/1704.01279
  • 博客:http://magenta.tensorflow.org/nsynth-instrument

图45:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Jesse Engelberg

自然语言处理

43、Learning to communicate: Agents developing their own language

  • 博客:http://blog.openai.com/learning-to-communicate/
  • 论文:http://arxiv.org/abs/1703.04908

图46:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:OpenAI

44、Big Picture Machine Learning: Classifying Text with Neural Networks and TensorFlow

  • GitHub:http://github.com/dmesquita/understanding_tensorflow_nn
  • 博客:http://medium.freecodecamp.org/big-picture-machine-learning-classifying-text-with-neural-networks-and-tensorflow-d94036ac2274

图47:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Déborah Mesquita

45、A novel approach to neural machine translation

  • GitHub:http://github.com/facebookresearch/fairseq
  • 论文:http://arxiv.org/abs/1705.03122
  • 博客:http://code.facebook.com/posts/1978007565818999/a-novel-approach-to-neural-machine-translation

图48:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源: Facebook

46、How to make a racist AI without really trying

  • Jupyter Python:http://gist.github.com/rspeer/ef750e7e407e04894cb3b78a82d66aed
  • 博客:http://blog.conceptnet.io/2017/07/13/how-to-make-a-racist-ai-without-really-trying/

图49:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Rob Speer

学习预测

47、Using Machine Learning to Predict Value of Homes On Airbnb

  • 博客:http://medium.com/airbnb-engineering/using-machine-learning-to-predict-value-of-homes-on-airbnb-9272d3d4739d
  • 中文:http://github.com/xitu/gold-miner/blob/master/TODO/using-machine-learning-to-predict-value-of-homes-on-airbnb.md

图50:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Robert Chang

48、Engineering Uncertainty Estimation in Neural Networks for Time Series Prediction at Uber

  • 论文:http://arxiv.org/abs/1709.01907
  • 博客:http://eng.uber.com/neural-networks-uncertainty-estimation/

图51:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Uber

49、Using Machine Learning to make parking easier

  • 博客:http://research.googleblog.com/2017/02/using-machine-learning-to-predict.html
  • 产品介绍:http://blog.google/products/maps/know-you-go-parking-difficulty-google-maps/

图52:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
来源:Google

50、How to Predict Stock Prices Easily — Intro to Deep Learning #7

  • 视频:http://www.youtube.com/watch?v=ftMq5ps503w
  • 说明:http://github.com/llSourcell/How-to-Predict-Stock-Prices-Easily
  • Demo:GitHub:http://github.com/erilyth/DeepLearning-Challenges/tree/master/Image_Classifier

图53:[译] 15 大领域、50 篇文章,2018 年应当这样学习机器学习
作者:Siraj Raval

原文链接:
http://github.com/Mybridge/learn-machine-learning
http://medium.mybridge.co/learn-to-build-a-machine-learning-application-from-top-articles-of-2017-cdd5638453fc

余下全文(1/3)
分享这篇文章:

请关注我们:

发表评论

电子邮件地址不会被公开。 必填项已用*标注