閱讀書籍 - Interpretable Machine Learning
A Guide for Making Black Box Models Explainable.
兩篇論文:
https://arxiv.org/pdf/2001.02478.pdf
https://arxiv.org/pdf/1909.03012.pdf
書籍網址 https://christophm.github.io/interpretable-ml-book/
倚任副理建議直接先閱讀第五章和第六章
讀書計畫初稿如下表,從下週二開始分享,之後視情況可機動調整
Hsin-Chih Kao changed 5 years agoView mode Like Bookmark
機器學習的模型是訓練數據的產出,刪除任一訓練數據會影響訓練結果。若刪除每一訓練數據對模型產生巨大影響,則稱這個點唯有影響的點(instance)。對有影響的點分析可以幫助我們檢視模型。
Deletion Diagnostics : delete the instance from the training data, retrain the model on the reduced training dataset and observe the difference in the model parameters or predictions
Influenced functions : upweight a data instance by approximating the parameter changes based on the gradients of the model parameters.
6.4.1 Deletion Diagnostics
DFBETA : 衡量移除某個instance對模型參數的影響。
$DFBETA_i = β-β^{-i}$
適用於有參數的模型,如 logistic regression or neural networks.
A counterfactual explanation of a prediction describes the smallest change to the feature values that changes the prediction to a predefined output.
How do we define a good counterfactual explanation?
counterfactual instance produces the predefined prediction as closely as possible
counterfactual should be as similar as possible to the instance regarding feature values
a counterfactual instance should have feature values that are likely
Hsin-Chih Kao changed 5 years agoView mode Like Bookmark
MMD-critic 如何做為解釋型 AI ?
MMD-critic 提供三面向的解釋-
better understand the data distribution
building an interpretable model
making a black box model interpretable
Prototype
A prototype is a data instance that is representative of all the data.
用一個特定的data instance代表母體
Hsin-Chih Kao changed 5 years agoView mode Like Bookmark
*大家蒐集到覺得有興趣的文獻
Explainable AI:From Theory to Mo7va7on, Applica7ons and Challenges
https://euads.org/wp-content/uploads/2019/09/xai_Lecture_12.09.2019-Giannotti_2.pdf
Limitations of Interpretable Machine Learning Methods
https://compstat-lmu.github.io/iml_methods_limitations/
pypychen changed 5 years agoView mode Like Bookmark
Theory:
LIME (Local Interpretable Model-Agnostic Explanations)
限制式
x : instance
f : original model
G : the family of possible explanations 可解釋的模型(如:線性回歸)
πx : 跟x的距離(相似度)
pypychen changed 5 years agoView mode Like Bookmark