Bandits for Learning to Explain from Explanations
arXiv:2102.03815
Abstract
We introduce Explearn, an online algorithm that learns to jointly output predictions and explanations for those predictions. Explearn leverages Gaussian Processes (GP)-based contextual bandits. This brings two key benefits. First, GPs naturally capture different kinds of explanations and enable the system designer to control how explanations generalize across the space by virtue of choosing a suitable kernel. Second, Explearn builds on recent results in contextual bandits which guarantee convergence with high probability. Our initial experiments hint at the promise of the approach.
arXiv commentAccepted at the Explainable Agency in Artificial Intelligence Workshop, hosted at the 35th AAAI Conference on Artificial Intelligence, February 2-9, 2021
References in corpus (8)
- Understanding Black-box Predictions via Influence Functions
- Towards Robust Interpretability with Self-Explaining Neural Networks
- Local Rule-Based Explanations of Black Box Decision Systems
- Interpreting Blackbox Models via Model Extraction
- Characteristic and Universal Tensor Product Kernels
- Exploiting compositionality to explore a large space of model structures
- A simpler condition for consistency of a kernel independence test
- Machine Guides, Human Supervises: Interactive Learning with Global Explanations