理解（interpret）表示用可被认知（understandable）的说法去解释（explain）或呈现（present）。在机器学习的场景中，可解释性（interpretability）就表示模型能够使用人类可认知的说法进行解释和呈现。[Finale Doshi-Velez]
机器学习模型被许多人称为“黑盒”。这意味着虽然我们可以从中获得准确的预测，但我们无法清楚地解释或识别这些预测背后的逻辑。但是我们如何从模型中提取重要的见解呢？要记住哪些事项以及我们需要实现哪些功能或工具？这些是在提出模型可解释性问题时会想到的重要问题。
所有资源下载地址，见源地址。
本资源含了近年来热门的可解释人工智能(XAI)的前沿研究。从下图我们可以看到可解释/可解释AI的趋势。关于这个主题的出版物正在蓬勃发展。
下图展示了XAI的几个用例。在这里，根据这个数字将出版物分成几个类别。
研究性论文
The elephant in the interpretability room: Why use attention as explanation when we have saliency methods, EMNLP Workshop 2020
Explainable Machine Learning in Deployment, FAT 2020
A brief survey of visualization methods for deep learning models from the perspective of Explainable AI, Information Visualization 2020
Explaining Explanations in AI, ACM FAT 2019
Machine learning interpretability: A survey on methods and metrics, Electronics, 2019
A Survey on Explainable Artificial Intelligence (XAI): Towards Medical XAI, IEEE TNNLS 2020
Interpretable machine learning: definitions, methods, and applications, Arxiv preprint 2019
Visual Analytics in Deep Learning: An Interrogative Survey for the Next Frontiers, IEEE Transactions on Visualization and Computer Graphics, 2019
Explainable Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI, Information Fusion, 2019
Evaluating Explanation Without Ground Truth in Interpretable Machine Learning, Arxiv preprint 2019
A survey of methods for explaining black box models, ACM Computing Surveys, 2018
Explaining Explanations: An Overview of Interpretability of Machine Learning, IEEE DSAA, 2018
Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI), IEEE Access, 2018
Explainable artificial intelligence: A survey, MIPRO, 2018
How Convolutional Neural Networks See the World — A Survey of Convolutional Neural Network Visualization Methods, Mathematical Foundations of Computing 2018
Explainable Artificial Intelligence: Understanding, Visualizing and Interpreting Deep Learning Models, Arxiv 2017
Towards A Rigorous Science of Interpretable Machine Learning, Arxiv preprint 2017
Explaining Explanation, Part 1: Theoretical Foundations, IEEE Intelligent System 2017
Explaining Explanation, Part 2: Empirical Foundations, IEEE Intelligent System 2017
Explaining Explanation, Part 3: The Causal Landscape, IEEE Intelligent System 2017
Explaining Explanation, Part 4: A Deep Dive on Deep Nets, IEEE Intelligent System 2017
An accurate comparison of methods for quantifying variable importance in artificial neural networks using simulated data, Ecological Modelling 2004
Review and comparison of methods to study the contribution of variables in artificial neural network models, Ecological Modelling 2003
书籍
Explainable Artificial Intelligence (xAI) Approaches and Deep Meta-Learning Models, Advances in Deep Learning Chapter 2020
Explainable AI: Interpreting, Explaining and Visualizing Deep Learning, Springer 2019
Explanation in Artificial Intelligence: Insights from the Social Sciences, 2017 arxiv preprint
Visualizations of Deep Neural Networks in Computer Vision: A Survey, Springer Transparent Data Mining for Big and Small Data 2017
Explanatory Model Analysis Explore, Explain and Examine Predictive Models
Interpretable Machine Learning A Guide for Making Black Box Models Explainable
An Introduction to Machine Learning Interpretability An Applied Perspective on Fairness, Accountability, Transparency,and Explainable AI
开源课程
Interpretability and Explainability in Machine Learning, Harvard University
文章
We mainly follow the taxonomy in the survey paper and divide the XAI/XML papers into the several branches.
1. Transparent Model Design
2. Post-Explanation
2.1 Model Explanation(Model-level)
2.2 Model Inspection
2.3 Outcome Explanation
2.3.1 Feature Attribution/Importance(Saliency Map)
2.4 Neuron Importance
2.5 Example-based Explanations
2.5.1 Counterfactual Explanations(Recourse)
2.5.2 Influential Instances
2.5.3 Prototypes&Criticisms
Uncategorized Papers on Model/Instance Explanation
Does Explainable Artificial Intelligence Improve Human Decision-Making?, AAAI 2021
Incorporating Interpretable Output Constraints in Bayesian Neural Networks, NeuIPS 2020
Towards Interpretable Natural Language Understanding with Explanations as Latent Variables, NeurIPS 2020
Learning identifiable and interpretable latent models of high-dimensional neural activity using pi-VAE, NeurIPS 2020
Generative causal explanations of black-box classifiers, NeurIPS 2020
Learning outside the Black-Box: The pursuit of interpretable models, NeurIPS 2020
Explaining Groups of Points in Low-Dimensional Representations, ICML 2020
Explaining Knowledge Distillation by Quantifying the Knowledge, CVPR 2020
Fanoos: Multi-Resolution, Multi-Strength, Interactive Explanations for Learned Systems, IJCAI 2020
Machine Learning Explainability for External Stakeholders, IJCAI 2020
Py-CIU: A Python Library for Explaining Machine Learning Predictions Using Contextual Importance and Utility, IJCAI 2020
Machine Learning Explainability for External Stakeholders, IJCAI 2020
Interpretable Models for Understanding Immersive Simulations, IJCAI 2020
Towards Automatic Concept-based Explanations, NIPS 2019
Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead, Nature Machine Intelligence 2019
Interpretml: A unified framework for machine learning interpretability, arxiv preprint 2019
All Models are Wrong, but Many are Useful: Learning a Variable’s Importance by Studying an Entire Class of Prediction Models Simultaneously, JMLR 2019
On the Robustness of Interpretability Methods, ICML 2018 workshop
Towards A Rigorous Science of Interpretable Machine Learning, Arxiv preprint 2017
Object Region Mining With Adversarial Erasing: A Simple Classification to Semantic Segmentation Approach, CVPR 2017
LOCO, Distribution-Free Predictive Inference For Regression, Arxiv preprint 2016
Explaining data-driven document classifications, MIS Quarterly 2014
评测方法
Evaluations and Methods for Explanation through Robustness Analysis, arxiv preprint 2020
Evaluating and Aggregating Feature-based Model Explanations, IJCAI 2020
Sanity Checks for Saliency Metrics, AAAI 2020
A benchmark for interpretability methods in deep neural networks, NIPS 2019
Methods for interpreting and understanding deep neural networks, Digital Signal Processing 2017
Evaluating the visualization of what a Deep Neural Network has learned, IEEE Transactions on Neural Networks and Learning Systems 2015
Python开源库
AIF360: https://github.com/Trusted-AI/AIF360,
AIX360: https://github.com/IBM/AIX360,
Anchor: https://github.com/marcotcr/anchor, scikit-learn
Alibi: https://github.com/SeldonIO/alibi
Alibi-detect: https://github.com/SeldonIO/alibi-detect
BlackBoxAuditing: https://github.com/algofairness/BlackBoxAuditing, scikit-learn
Boruta-Shap: https://github.com/Ekeany/Boruta-Shap, scikit-learn
casme: https://github.com/kondiz/casme, Pytorch
Captum: https://github.com/pytorch/captum, Pytorch,
cnn-exposed: https://github.com/idealo/cnn-exposed, Tensorflow
DALEX: https://github.com/ModelOriented/DALEX,
Deeplift: https://github.com/kundajelab/deeplift, Tensorflow, Keras
DeepExplain: https://github.com/marcoancona/DeepExplain, Tensorflow, Keras
Deep Visualization Toolbox: https://github.com/yosinski/deep-visualization-toolbox, Caffe,
Eli5: https://github.com/TeamHG-Memex/eli5, Scikit-learn, Keras, xgboost, lightGBM, catboost etc.
explainx: https://github.com/explainX/explainx, xgboost, catboost
Grad-cam-Tensorflow: https://github.com/insikk/Grad-CAM-tensorflow, Tensorflow
Innvestigate: https://github.com/albermax/innvestigate, tensorflow, theano, cntk, Keras
imodels: https://github.com/csinva/imodels,
InterpretML: https://github.com/interpretml/interpret
interpret-community: https://github.com/interpretml/interpret-community
Integrated-Gradients: https://github.com/ankurtaly/Integrated-Gradients, Tensorflow
Keras-grad-cam: https://github.com/jacobgil/keras-grad-cam, Keras
Keras-vis: https://github.com/raghakot/keras-vis, Keras
keract: https://github.com/philipperemy/keract, Keras
Lucid: https://github.com/tensorflow/lucid, Tensorflow
LIT: https://github.com/PAIR-code/lit, Tensorflow, specified for NLP Task
Lime: https://github.com/marcotcr/lime, Nearly all platform on Python
LOFO: https://github.com/aerdem4/lofo-importance, scikit-learn
modelStudio: https://github.com/ModelOriented/modelStudio, Keras, Tensorflow, xgboost, lightgbm, h2o
pytorch-cnn-visualizations: https://github.com/utkuozbulak/pytorch-cnn-visualizations, Pytorch
Pytorch-grad-cam: https://github.com/jacobgil/pytorch-grad-cam, Pytorch
PDPbox: https://github.com/SauceCat/PDPbox, Scikit-learn
py-ciu:https://github.com/TimKam/py-ciu/,
PyCEbox: https://github.com/AustinRochford/PyCEbox
path_explain: https://github.com/suinleelab/path_explain, Tensorflow
rulefit: https://github.com/christophM/rulefit,
rulematrix: https://github.com/rulematrix/rule-matrix-py,
Saliency: https://github.com/PAIR-code/saliency, Tensorflow
SHAP: https://github.com/slundberg/shap, Nearly all platform on Python
Skater: https://github.com/oracle/Skater
TCAV: https://github.com/tensorflow/tcav, Tensorflow, scikit-learn
skope-rules: https://github.com/scikit-learn-contrib/skope-rules, Scikit-learn
TensorWatch: https://github.com/microsoft/tensorwatch.git, Tensorflow
tf-explain: https://github.com/sicara/tf-explain, Tensorflow
Treeinterpreter: https://github.com/andosa/treeinterpreter, scikit-learn,
WeightWatcher: https://github.com/CalculatedContent/WeightWatcher, Keras, Pytorch
What-if-tool: https://github.com/PAIR-code/what-if-tool, Tensorflow
XAI: https://github.com/EthicalML/xai, scikit-learn
Related Repositories
https://github.com/jphall663/awesome-machine-learning-interpretability,
https://github.com/lopusz/awesome-interpretable-machine-learning,
