没有合适的资源?快使用搜索试试~ 我知道了~
首页Interpretable Machine Learning by Christoph Molnar .pdf
资源详情
资源推荐
Interpretable Machine Learning
Interpretable Machine Learning
A Guide for Making Black Box Models Explainable.
Christoph Molnar
2021-03-12
Summary
Machine learning has great potential for improving products, processes and research. But computers usually do not
explain their predictions which is a barrier to the adoption of machine learning. This book is about making machine
learning models and their decisions interpretable.
After exploring the concepts of interpretability, you will learn about simple, interpretable models such as decision
trees, decision rules and linear regression. Later chapters focus on general model-agnostic methods for interpreting
black box models like feature importance and accumulated local effects and explaining individual predictions with
Shapley values and LIME.
All interpretation methods are explained in depth and discussed critically. How do they work under the hood? What
are their strengths and weaknesses? How can their outputs be interpreted? This book will enable you to select and
correctly apply the interpretation method that is most suitable for your machine learning project.
The book focuses on machine learning models for tabular data (also called relational or structured data) and less on
computer vision and natural language processing tasks. Reading the book is recommended for machine learning
practitioners, data scientists, statisticians, and anyone else interested in making machine learning models interpretable.
You can buy the PDF and e-book version (epub, mobi) on leanpub.com.
You can buy the print version on lulu.com.
About me: My name is Christoph Molnar, I'm a statistician and a machine learner. My goal is to make machine
learning interpretable.
Mail: christoph.molnar.ai@gmail.com
Website: https://christophm.github.io/
Follow me on Twitter! @ChristophMolnar
Cover by @YvonneDoinel
Creative Commons License
This book is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Interpretable machine learning•
• Summary•
Preface by the Author•
1 Introduction
1.1 Story Time
Lightning Never Strikes Twice◊
Trust Fall◊
Fermi's Paperclips◊
♦
1.2 What Is Machine Learning?♦
1.3 Terminology♦
•
2 Interpretability
2.1 Importance of Interpretability♦
2.2 Taxonomy of Interpretability Methods♦
2.3 Scope of Interpretability
2.3.1 Algorithm Transparency◊
2.3.2 Global, Holistic Model Interpretability◊
2.3.3 Global Model Interpretability on a Modular Level◊
2.3.4 Local Interpretability for a Single Prediction◊
2.3.5 Local Interpretability for a Group of Predictions◊
♦
2.4 Evaluation of Interpretability♦
2.5 Properties of Explanations♦
2.6 Human-friendly Explanations
2.6.1 What Is an Explanation?◊
2.6.2 What Is a Good Explanation?◊
♦
•
3 Datasets
3.1 Bike Rentals (Regression)♦
3.2 YouTube Spam Comments (Text Classification)♦
3.3 Risk Factors for Cervical Cancer (Classification)♦
•
4 Interpretable Models
4.1 Linear Regression
4.1.1 Interpretation◊
4.1.2 Example◊
4.1.3 Visual Interpretation◊
4.1.4 Explain Individual Predictions◊
4.1.5 Encoding of Categorical Features◊
4.1.6 Do Linear Models Create Good Explanations?◊
4.1.7 Sparse Linear Models◊
4.1.8 Advantages◊
4.1.9 Disadvantages◊
♦
4.2 Logistic Regression
4.2.1 What is Wrong with Linear Regression for Classification?◊
♦
•
4.2.2 Theory◊
4.2.3 Interpretation◊
4.2.4 Example◊
4.2.5 Advantages and Disadvantages◊
4.2.6 Software◊
4.3 GLM, GAM and more
4.3.1 Non-Gaussian Outcomes - GLMs◊
4.3.2 Interactions◊
4.3.3 Nonlinear Effects - GAMs◊
4.3.4 Advantages◊
4.3.5 Disadvantages◊
4.3.6 Software◊
4.3.7 Further Extensions◊
♦
4.4 Decision Tree
4.4.1 Interpretation◊
4.4.2 Example◊
4.4.3 Advantages◊
4.4.4 Disadvantages◊
4.4.5 Software◊
♦
4.5 Decision Rules
4.5.1 Learn Rules from a Single Feature (OneR)◊
4.5.2 Sequential Covering◊
4.5.3 Bayesian Rule Lists◊
4.5.4 Advantages◊
4.5.5 Disadvantages◊
4.5.6 Software and Alternatives◊
♦
4.6 RuleFit
4.6.1 Interpretation and Example◊
4.6.2 Theory◊
4.6.3 Advantages◊
4.6.4 Disadvantages◊
4.6.5 Software and Alternative◊
♦
4.7 Other Interpretable Models
4.7.1 Naive Bayes Classifier◊
4.7.2 K-Nearest Neighbors◊
♦
5 Model-Agnostic Methods
5.1 Partial Dependence Plot (PDP)
5.1.1 Examples◊
5.1.2 Advantages◊
5.1.3 Disadvantages◊
5.1.4 Software and Alternatives◊
♦
5.2 Individual Conditional Expectation (ICE)
5.2.1 Examples◊
5.2.2 Advantages◊
5.2.3 Disadvantages◊
5.2.4 Software and Alternatives◊
♦
5.3 Accumulated Local Effects (ALE) Plot
5.3.1 Motivation and Intuition◊
5.3.2 Theory◊
5.3.3 Estimation◊
5.3.4 Examples◊
5.3.5 Advantages◊
5.3.6 Disadvantages◊
5.3.7 Implementation and Alternatives◊
♦
5.4 Feature Interaction
5.4.1 Feature Interaction?◊
5.4.2 Theory: Friedman's H-statistic◊
5.4.3 Examples◊
♦
•
5.4.4 Advantages◊
5.4.5 Disadvantages◊
5.4.6 Implementations◊
5.4.7 Alternatives◊
5.5 Permutation Feature Importance
5.5.1 Theory◊
5.5.2 Should I Compute Importance on Training or Test Data?◊
5.5.3 Example and Interpretation◊
5.5.4 Advantages◊
5.5.5 Disadvantages◊
5.5.6 Software and Alternatives◊
♦
5.6 Global Surrogate
5.6.1 Theory◊
5.6.2 Example◊
5.6.3 Advantages◊
5.6.4 Disadvantages◊
5.6.5 Software◊
♦
5.7 Local Surrogate (LIME)
5.7.1 LIME for Tabular Data◊
5.7.2 LIME for Text◊
5.7.3 LIME for Images◊
5.7.4 Advantages◊
5.7.5 Disadvantages◊
♦
5.8 Scoped Rules (Anchors)
5.8.1 Finding Anchors◊
5.8.2 Complexity and Runtime◊
5.8.3 Tabular Data Example◊
5.8.4 Advantages◊
5.8.5 Disadvantages◊
5.8.6 Software and Alternatives◊
♦
5.9 Shapley Values
5.9.1 General Idea◊
5.9.2 Examples and Interpretation◊
5.9.3 The Shapley Value in Detail◊
5.9.4 Advantages◊
5.9.5 Disadvantages◊
5.9.6 Software and Alternatives◊
♦
5.10 SHAP (SHapley Additive exPlanations)
5.10.1 Definition◊
5.10.2 KernelSHAP◊
5.10.3 TreeSHAP◊
5.10.4 Examples◊
5.10.5 SHAP Feature Importance◊
5.10.6 SHAP Summary Plot◊
5.10.7 SHAP Dependence Plot◊
5.10.8 SHAP Interaction Values◊
5.10.9 Clustering SHAP values◊
5.10.10 Advantages◊
5.10.11 Disadvantages◊
5.10.12 Software◊
♦
6 Example-Based Explanations
6.1 Counterfactual Explanations
6.1.1 Generating Counterfactual Explanations◊
6.1.2 Example◊
6.1.3 Advantages◊
6.1.4 Disadvantages◊
6.1.5 Software and Alternatives◊
♦
6.2 Adversarial Examples♦
•
6.2.1 Methods and Examples◊
6.2.2 The Cybersecurity Perspective◊
6.3 Prototypes and Criticisms
6.3.1 Theory◊
6.3.2 Examples◊
6.3.3 Advantages◊
6.3.4 Disadvantages◊
6.3.5 Code and Alternatives◊
♦
6.4 Influential Instances
6.4.1 Deletion Diagnostics◊
6.4.2 Influence Functions◊
6.4.3 Advantages of Identifying Influential Instances◊
6.4.4 Disadvantages of Identifying Influential Instances◊
6.4.5 Software and Alternatives◊
♦
7 Neural Network Interpretation
7.1 Learned Features
7.1.1 Feature Visualization◊
7.1.2 Network Dissection◊
7.1.3 Advantages◊
7.1.4 Disadvantages◊
7.1.5 Software and Further Material◊
♦
7.2 Pixel Attribution (Saliency Maps)
7.2.1 Vanilla Gradient (Saliency Maps)◊
7.2.2 DeconvNet◊
7.2.3 Grad-CAM◊
7.2.4 Guided Grad-CAM◊
7.2.5 SmoothGrad◊
7.2.6 Examples◊
7.2.7 Advantages◊
7.2.8 Disadvantages◊
7.2.9 Software◊
♦
•
8 A Look into the Crystal Ball
8.1 The Future of Machine Learning♦
8.2 The Future of Interpretability♦
•
9 Contribute to the Book•
10 Citing this Book•
11 Translations•
12 Acknowledgements•
References
R Packages Used for Examples♦
•
• Published with bookdown•
剩余254页未读,继续阅读
m0_49448618
- 粉丝: 0
- 资源: 1
上传资源 快速赚钱
- 我的内容管理 收起
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
会员权益专享
最新资源
- zigbee-cluster-library-specification
- JSBSim Reference Manual
- c++校园超市商品信息管理系统课程设计说明书(含源代码) (2).pdf
- 建筑供配电系统相关课件.pptx
- 企业管理规章制度及管理模式.doc
- vb打开摄像头.doc
- 云计算-可信计算中认证协议改进方案.pdf
- [详细完整版]单片机编程4.ppt
- c语言常用算法.pdf
- c++经典程序代码大全.pdf
- 单片机数字时钟资料.doc
- 11项目管理前沿1.0.pptx
- 基于ssm的“魅力”繁峙宣传网站的设计与实现论文.doc
- 智慧交通综合解决方案.pptx
- 建筑防潮设计-PowerPointPresentati.pptx
- SPC统计过程控制程序.pptx
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功