A Generic Approach for Reproducible Model Distillation
This work addresses the need for reproducible and reliable interpretations in model distillation for users of interpretable machine learning, though it is incremental as it builds on existing stabilization methods by generalizing them.
The paper tackles the problem of unreliable interpretations in model distillation due to student model sensitivity to training data variability, proposing a generic approach based on the central limit theorem and multiple testing to select a corpus size for stable distillation, demonstrated on decision trees, falling rule lists, and symbolic regression with simulation experiments on Mammographic Mass and Breast Cancer datasets.
Model distillation has been a popular method for producing interpretable machine learning. It uses an interpretable "student" model to mimic the predictions made by the black box "teacher" model. However, when the student model is sensitive to the variability of the data sets used for training even when keeping the teacher fixed, the corresponded interpretation is not reliable. Existing strategies stabilize model distillation by checking whether a large enough corpus of pseudo-data is generated to reliably reproduce student models, but methods to do so have so far been developed for a specific student model. In this paper, we develop a generic approach for stable model distillation based on central limit theorem for the average loss. We start with a collection of candidate student models and search for candidates that reasonably agree with the teacher. Then we construct a multiple testing framework to select a corpus size such that the consistent student model would be selected under different pseudo samples. We demonstrate the application of our proposed approach on three commonly used intelligible models: decision trees, falling rule lists and symbolic regression. Finally, we conduct simulation experiments on Mammographic Mass and Breast Cancer datasets and illustrate the testing procedure throughout a theoretical analysis with Markov process. The code is publicly available at https://github.com/yunzhe-zhou/GenericDistillation.