Summary of Interpretcc: Intrinsic User-centric Interpretability Through Global Mixture Of Experts, by Vinitra Swamy et al.
InterpretCC: Intrinsic User-Centric Interpretability through Global Mixture of Experts
by Vinitra Swamy, Syrielle Montariol, Julian Blackwell, Jibril Frej, Martin Jaggi, Tanja Käser
First submitted to arxiv on: 5 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Most neural network interpretability methods compromise one or more of faithfulness, understandability, and model performance. To address this trade-off, we propose InterpretCC, a family of interpretable-by-design neural networks that guarantee human-centric interpretability while maintaining comparable performance to state-of-the-art models. Our approach adaptively and sparsely activates features before prediction, ensuring both faithfulness and understandability. We also extend this idea into an interpretable global mixture-of-experts (MoE) model that allows humans to specify topics of interest and adaptively activate topical subnetworks for prediction. We demonstrate the effectiveness of our approach on real-world benchmarks across text, time series, and tabular data. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Interpretability is important because it helps people understand why a neural network made a certain decision. Most methods that try to do this compromise one or more things, like how well they explain the decision or how easy it is for humans to understand the explanation. This can be a problem when we want to use these networks in situations where we need to trust the explanations, such as education and healthcare. Our new approach, called InterpretCC, tries to balance these different requirements by making sure that neural networks are not only good at explaining their decisions but also easy for humans to understand and accurate. |
Keywords
* Artificial intelligence * Mixture of experts * Neural network * Time series
