MEME: Generating RNN Model Explanations via Model Extraction
This work addresses the problem of improving the explainability and interpretability of RNNs for researchers and practitioners working with sequential data.
This paper introduces MEME, a model extraction approach that approximates Recurrent Neural Networks (RNNs) with interpretable models. The extracted models, represented by human-understandable concepts and their interactions, can be used to interpret RNN decision-making both locally and globally.
Recurrent Neural Networks (RNNs) have achieved remarkable performance on a range of tasks. A key step to further empowering RNN-based approaches is improving their explainability and interpretability. In this work we present MEME: a model extraction approach capable of approximating RNNs with interpretable models represented by human-understandable concepts and their interactions. We demonstrate how MEME can be applied to two multivariate, continuous data case studies: Room Occupation Prediction, and In-Hospital Mortality Prediction. Using these case-studies, we show how our extracted models can be used to interpret RNNs both locally and globally, by approximating RNN decision-making via interpretable concept interactions.