Granger-causal Attentive Mixtures of Experts: Learning Important Features with Neural Networks

Abstract

Knowledge of the importance of input features towards decisions made by machine-learning models is essential to increase our understanding of both the models and the underlying data. Here, we present a new approach to estimating feature importance with neural networks based on the idea of distributing the features of interest among experts in an attentive mixture of experts (AME). AMEs use attentive gating networks trained with a Granger-causal objective to learn to jointly produce accurate predictions as well as estimates of feature importance in a single model. Our experiments show (i) that the feature importance estimates provided by AMEs compare favourably to those provided by state-of-the-art methods, (ii) that AMEs are significantly faster at estimating feature importance than existing methods, and (iii) that the associations discovered by AMEs are consistent with those reported by domain experts.

arXiv commentAAAI Conference on Artificial Intelligence 2019

References in corpus (13)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Neural Machine Translation by Jointly Learning to Align and Translate
• A Unified Approach to Interpreting Model Predictions
• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
• Learning Important Features Through Propagating Activation Differences
• Explaining NonLinear Classification Decisions with Deep Taylor Decomposition
• Understanding Black-box Predictions via Influence Functions
• Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
• How to Explain Individual Classification Decisions
• Reasoning about Entailment with Neural Attention
• RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism
• Deep Networks with Internal Selective Attention through Feedback Connections
• Not to Cry Wolf: Distantly Supervised Multitask Learning in Critical Care

Cited by in corpus (7)

• Variable-lag Granger Causality and Transfer Entropy for Time Series Analysis
• A Deep Learning Approach to Diagnosing Multiple Sclerosis from Smartphone Data
• Variable-lag Granger Causality for Time Series Analysis
• Gaussian Process Regression with Local Explanation
• Neural Generators of Sparse Local Linear Models for Achieving both Accuracy and Interpretability
• Learning Channel Importance for High Content Imaging with Interpretable Deep Input Channel Mixing
• Towards a Universal Gating Network for Mixtures of Experts