Can I trust you more? Model-Agnostic Hierarchical Explanations

Abstract

Interactions such as double negation in sentences and scene interactions in images are common forms of complex dependencies captured by state-of-the-art machine learning models. We propose Mahé, a novel approach to provide Model-agnostic hierarchical éxplanations of how powerful machine learning models, such as deep neural networks, capture these interactions as either dependent on or free of the context of data instances. Specifically, Mahé provides context-dependent explanations by a novel local interpretation algorithm that effectively captures any-order interactions, and obtains context-free explanations through generalizing context-dependent interactions to explain global behaviors. Experimental results show that Mahé obtains improved local interaction interpretations over state-of-the-art methods and successfully explains interactions that are context-free.

References in corpus (8)

• SmoothGrad: removing noise by adding noise
• Consistent Individualized Feature Attribution for Tree Ensembles
• Pointer Sentinel Mixture Models
• Beyond Word Importance: Contextual Decomposition to Extract Interactions from LSTMs
• Hierarchical interpretations for neural network predictions
• Locally Interpretable Models and Effects based on Supervised Partitioning (LIME-SUP)
• Detecting Statistical Interactions from Neural Network Weights
• Efficient Representation of Low-Dimensional Manifolds using Deep Networks

Cited by in corpus (8)

• Interpretable machine learning: definitions, methods, and applications
• A One-Class Classification Decision Tree Based on Kernel Density Estimation
• How does this interaction affect me? Interpretable attribution for feature interactions
• Model Agnostic Multilevel Explanations
• ProtoryNet - Interpretable Text Classification Via Prototype Trajectories
• Explaining Neural Network Predictions on Sentence Pairs via Learning Word-Group Masks
• When Stability meets Sufficiency: Informative Explanations that do not Overwhelm
• Interpretable Artificial Intelligence through the Lens of Feature Interaction