Machine learning explainability via microaggregation and shallow decision trees
作者：
Highlights：
• We give explanations of deep learning decisions using shallow decision trees.
• Decision trees are computed on clusters obtained via microaggregation.
• The cluster size trades off comprehensibility, representativeness and privacy.
• We present experiments on large numerical and categorical data sets.
• For categorical data sets, we use ontologies for semantic consistency.
摘要
•We give explanations of deep learning decisions using shallow decision trees.•Decision trees are computed on clusters obtained via microaggregation.•The cluster size trades off comprehensibility, representativeness and privacy.•We present experiments on large numerical and categorical data sets.•For categorical data sets, we use ontologies for semantic consistency.
论文关键词：Explainability,Machine learning,Data protection,Microaggregation,Privacy
论文评审过程：Received 18 July 2019, Revised 13 January 2020, Accepted 16 January 2020, Available online 24 January 2020, Version of Record 18 May 2020.
论文官网地址：https://doi.org/10.1016/j.knosys.2020.105532
