Gaining Free or Low-Cost Interpretability with Interpretable Partial Substitute

By Tong Wang. 2019

48 citations recorded by the source corpus.

Subfields: Natural Language Processing, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

1. Satoshi Hara
2. Danqing Pan
3. Jingyi Yang
4. Yunyi Li
5. Boxiang Wang
6. Tong Wang
7. Sébastien Gambs
8. Zhi Chen

In Safety, Robustness & Interpretability

1. Lesia Semenova
2. Tong Wang
3. Satoshi Hara
4. Zhi Chen
5. Danqing Pan
6. Chudi Zhong
7. Haiyang Huang
8. Cynthia Rudin

Citations in the corpus

• Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (2021, background): “[309, 312], it is possible that easier cases could be handled by simpler models, leaving only the harder cases for models with less interpretability.”
• Partially Interpretable Estimators (PIE): Black-Box-Refined Interpretable Machine Learning (2021, methodology): “Model Hybrids The idea of utilizing an interpretable model and a black-box model jointly to make a prediction is similar to recent works of hybrid models (Wang, 2019; Pan et al., 2020).”
• Learning Fair Rule Lists (2019, background): “More recently, an hybrid framework for transparency (Wang, 2019; Rafique et al., 2019) has been proposed to associate black-box models with their interpretable partial substitutes.”
• A Survey on Neural Network Interpretability (2020, background): “[48] attempted to find an interpretable partial substitution (a rule set) to the network that covers a certain subset of the input space.”
• Interpretable Companions for Black-Box Models (2020, methodology): “Despite the similarity, we would like to emphasize that HRS is not flexible.”
• A Semiparametric Approach to Interpretable Machine Learning (2020, unknown): “mple, [2] combined the interpretability of coefficients in a linear regression model with the model flexibility of neural networks to yield greater performance while preserving interpretability, whilst [41] learned an interpretable model in place of a black-box model for subsets of the data, while lea”