# 5.6 Global Surrogate ---------------- ## Theory: * 必須是機器學習模型 * 具可解釋性 * 訓練目標是Black Box模型的“預測值”,而非真實資料的“目標值” * 訓練資料可以是: * 與Black Box模型的訓練資料相同 * 具同樣分布的新資料集 * 原資料的Subset * 衡量代理模型成效:R square (線性模型) ## Advantages: * 有彈性 * 直覺 ## Disadvantages: * R^2的切點不明確 * 代理模型可能會產生局部資料適用,局部資料完全不適用的狀況 參考資料: [https://smt.readthedocs.io/en/latest/](https://smt.readthedocs.io/en/latest/) [https://github.com/SMTorg/smt](https://github.com/SMTorg/smt) [https://en.wikipedia.org/wiki/Surrogate_model](https://en.wikipedia.org/wiki/Surrogate_model) 其他代理模型清單: CH 4.1 所提到的 [linear regression](https://christophm.github.io/interpretable-ml-book/limo.html#limo), [logistic regression](https://christophm.github.io/interpretable-ml-book/logistic.html#logistic), [other linear regression extensions](https://christophm.github.io/interpretable-ml-book/extend-lm.html#extend-lm), [decision trees](https://christophm.github.io/interpretable-ml-book/tree.html#tree), [decision rules](https://christophm.github.io/interpretable-ml-book/rules.html#rules) and [the RuleFit algorithm](https://christophm.github.io/interpretable-ml-book/rulefit.html#rulefit) ###### tags: `重點摘要`
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up