ExtraTreeRegressor over IsolationTree in IsolationForest #27918
Unanswered
albertoazzari
asked this question in
Q&A
Replies: 1 comment 1 reply
-
This is following the proposal of the original paper: http://www.lamda.nju.edu.cn/publication/icdm08b.pdf |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I have been exploring the implementation of Isolation Forest and noticed that the trees within the Isolation Forest are based on the Extra Tree Regressor instead of the conventional Isolation Tree.
I'm curious about the reasoning behind this design choice. Could you please provide insights into why the Extra Tree Regressor (with a split optimization on a random vector y) was chosen over a standard Isolation Tree for the Isolation Forest?
Additionally, I'm interested in understanding how the optimization of splits with the squared_error criterion in the Extra Tree Regressor influences the performance of the Isolation Forest. Does this optimization have a positive impact on the predictive capabilities of the Isolation Forest, and if so, in what ways?
I appreciate your time and expertise in clarifying these aspects. Thank you for your contribution to scikit-learn, and I look forward to gaining a deeper understanding of these design decisions.
Best regards,
Alberto
Beta Was this translation helpful? Give feedback.
All reactions