AI實戰聖經《Machine Learning Yearning》第1-52章中英文版pdf分享

    《Machine Learning Yearning》是機器學習泰斗Andrew NG花了近2年時間，根據自己多年實踐經驗整理出來的一本機器學習、深度學習實踐經驗寶典。本書的重點不在於教授傳統的機器學習演算法理論基礎，而在於教你如何在實踐中使機器學習演演算法的實戰經驗。如果你渴望成為AI的技術領導者，並想要學習如何為團隊設定一個方向，本書將有所幫助。

 本書官方網址：http://www.mlyearning.org/

    臺主花了幾天時間對本書1-52節的中英文內容進行了整理，內容整理自網路。文末附本書中文和英文pdf下載地址，僅供學習分享。

    本書主要總結了50多個吳恩達多年在AI領域的工程要領，把每一個要領都濃縮到 1-2 頁的閱讀量，非常精煉。目前，前52個要領已經分享出來了，被分為9個主題。

前9個主題串列

    第一章：緒論 「Introduction」

    第二章：配置開發集和訓練集 「Setting up development and test sets」

    第三章：基本誤差分析 「Basic Error Analysis」

    第四章：偏差和方差 「Bias and Variance」

    第五章：學習曲線 「Learning curves」

    第六章：比較人類水平表現 「Comparing to human-level performance」

    第七章：不同分佈下的訓練和測試 「Training and testing on different distributions」

    第八章：除錯推理演演算法 「Debugging inference algorithms」

    第九章：端到端的深度學習 「End-to-end deep learning」

前52個要領串列

（英文串列，保證原汁原味）

    1 Why Machine Learning Strategy

    2 How to use this book to help your team

    3 Prerequisites and Notation

    4 Scale drives machine learning progress

    5 Your development and test sets

    6 Your dev and test sets should come from the same distribution

    7 How large do the dev/test sets need to be?

    8 Establish a single-number evaluation metric for your team to optimize

    9 Optimizing and satisficing metrics

    10 Having a dev set and metric speeds up iterations

    11 When to change dev/test sets and metrics

    12 Takeaways: Setting up development and test sets

    13 Build your first system quickly, then iterate

    14 Error analysis: Look at dev set examples to evaluate ideas

    15 Evaluating multiple ideas in parallel during error analysis

    16 Cleaning up mislabeled dev and test set examples

    17 If you have a large dev set, split it into two subsets, only one of which you look at

    18 How big should the Eyeball and Blackbox dev sets be?

    19 Takeaways: Basic error analysis

    20 Bias and Variance: The two big sources of error

    21 Examples of Bias and Variance

    22 Comparing to the optimal error rate

    23 Addressing Bias and Variance

    24 Bias vs. Variance tradeoff

    25 Techniques for reducing avoidable bias

    Page 3 Machine Learning Yearning-Draft Andrew Ng26 Techniques for reducing Variance

    27 Error analysis on the training set

    28 Diagnosing bias and variance: Learning curves

    29 Plotting training error

    30 Interpreting learning curves: High bias

    31 Interpreting learning curves: Other cases

    32 Plotting learning curves

    33 Why we compare to human-level performance

    34 How to define human-level performance

    35 Surpassing human-level performance

    36 Why train and test on different distributions

    37 Whether to use all your data

    38 Whether to include inconsistent data

    39 Weighting data

    40 Generalizing from the training set to the dev set

    41 Addressing Bias, and Variance, and Data Mismatch

    42 Addressing data mismatch

    43 Artificial data synthesis

    44 The Optimization Verification test

    45 General form of Optimization Verification test

    46 Reinforcement learning example

    47 The rise of end-to-end learning

    48 More end-to-end learning examples

    49 Pros and cons of end-to-end learning

    50 Learned sub-components

    51 Directly learning rich outputs

    52 Error Analysis by Parts