Granger-causal Attentive Mixtures of Experts: Learning Important Features with Neural Networks
On the map: Mixture-of-experts → MoE theory and gating functions
Abstract (arXiv)
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.
Key facts. Cited by 13 papers in the map (data through 7 Oct 2026). Placed in MoE theory and gating functions, one of 12 directions in Mixture-of-experts. Builds on 31 of its 73 references that are in the map. First posted on arXiv 6 Feb 2018.
Builds on
31 of 73 references.
References · 31
- 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 6,979
- 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,690
- 2017A Unified Approach to Interpreting Model Predictionscited by 6,488
- 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
- 2014Neural Machine Translation by Jointly Learning to Align and Translatecited by 4,530
- 2017Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layercited by 3,413
- 2017Axiomatic Attribution for Deep Networkscited by 3,350
- 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
- 2013Visualizing and Understanding Convolutional Networkscited by 2,278
- 2016Equality of Opportunity in Supervised Learningcited by 2,000
- 2017Understanding Black-box Predictions via Influence Functionscited by 1,714
- 2017Learning Important Features Through Propagating Activation Differencescited by 1,454
19 more
- 2015Show, Attend and Tell: Neural Image Caption Generation with Visual Attentioncited by 1,421
- 2017Towards A Rigorous Science of Interpretable Machine Learningcited by 1,400
- 2017SmoothGrad: removing noise by adding noisecited by 1,097
- 2017Interpretable Explanations of Black Boxes by Meaningful Perturbationcited by 611
- 2016RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanismcited by 344
- 2016European Union regulations on algorithmic decision-making and a "right to explanation"cited by 343
- 2017The (Un)reliability of saliency methodscited by 331
- 2015Explaining NonLinear Classification Decisions with Deep Taylor Decompositioncited by 248
- 2017Distral: Robust Multitask Reinforcement Learningcited by 231
- 2016Model-Agnostic Interpretability of Machine Learningcited by 219
- 2017Multi-task Self-Supervised Visual Learningcited by 200
- 2015Conditional Computation in Neural Networks for faster modelscited by 166
- 2015Reasoning about Entailment with Neural Attentioncited by 68
- 2017Beyond Sparsity: Tree Regularization of Deep Models for Interpretabilitycited by 68
- 2017MDNet: A Semantically and Visually Interpretable Medical Image Diagnosis Networkcited by 56
- 2016Auditing Black-box Models for Indirect Influencecited by 41
- 2014Deep Networks with Internal Selective Attention through Feedback Connectionscited by 17
- 2017Classification of Radiology Reports Using Neural Attention Modelscited by 5
- 2017Beat by Beat: Classifying Cardiac Arrhythmias with Recurrent Neural Networkscited by 5
Built on by
Citing papers per year: 2019: 1; 2020: 1; 2021: 4; 2022: 0; 2023: 4; 2024: 2; 2025: 1.
13 papers.
Sort:
By area:
- 2023Causal Discovery from Temporal Data: An Overview and New Perspectivescited by 43
- 2023Neural Markov Jump Processescited by 11
- 2021Inductive Granger Causal Modeling for Multivariate Time Seriescited by 8
- 2021Instance-wise Causal Feature Selection for Model Interpretationcited by 8
- 2021Temporal-Spatial Causal Interpretations for Vision-Based Reinforcement Learningcited by 6
- 2019A Modular Task-oriented Dialogue System Using a Neural Mixture-of-Expertscited by 4
- 2023Learning Channel Importance for High Content Imaging with Interpretable Deep Input Channel Mixingcited by 3
- 2024Causal Feature Selection for Responsible Machine Learningcited by 2
- 2021Training Deep Models to be Explained with Fewer Examplescited by 1
- 2024Linking Model Intervention to Causal Interpretation in Model Explanationcited by 1
- 2025TPK: Trustworthy Trajectory Prediction Integrating Prior Knowledge For Interpretability and Kinematic Feasibilitycited by 1
- 2020Towards a Universal Gating Network for Mixtures of Expertscited by 0
- 2025TPK: Trustworthy Trajectory Prediction Integrating Prior Knowledge For Interpretability and Kinematic Feasibility10 May 2025
- 2024Linking Model Intervention to Causal Interpretation in Model Explanation21 Oct 2024
- 2024Causal Feature Selection for Responsible Machine Learning5 Feb 2024
- 2023Explanation-based Training with Differentiable Insertion/Deletion Metric-aware Regularizers19 Oct 2023
- 2023Learning Channel Importance for High Content Imaging with Interpretable Deep Input Channel Mixing31 Aug 2023
- 2023Neural Markov Jump Processes31 May 2023
- 2023Causal Discovery from Temporal Data: An Overview and New Perspectives17 Mar 2023
- 2021Training Deep Models to be Explained with Fewer Examples7 Dec 2021
- 2021Temporal-Spatial Causal Interpretations for Vision-Based Reinforcement Learning6 Dec 2021
- 2021Instance-wise Causal Feature Selection for Model Interpretation26 Apr 2021
- 2021Inductive Granger Causal Modeling for Multivariate Time Series10 Feb 2021
- 2020Towards a Universal Gating Network for Mixtures of Experts3 Nov 2020
1 more
- 2023Causal Discovery from Temporal Data: An Overview and New Perspectivescited by 43
- 2023Neural Markov Jump Processescited by 11
- 2021Inductive Granger Causal Modeling for Multivariate Time Seriescited by 8
- 2021Instance-wise Causal Feature Selection for Model Interpretationcited by 8
- 2021Temporal-Spatial Causal Interpretations for Vision-Based Reinforcement Learningcited by 6
- 2023Learning Channel Importance for High Content Imaging with Interpretable Deep Input Channel Mixingcited by 3
- 2024Causal Feature Selection for Responsible Machine Learningcited by 2
- 2021Training Deep Models to be Explained with Fewer Examplescited by 1
- 2024Linking Model Intervention to Causal Interpretation in Model Explanationcited by 1
- 2025TPK: Trustworthy Trajectory Prediction Integrating Prior Knowledge For Interpretability and Kinematic Feasibilitycited by 1
- 2023Explanation-based Training with Differentiable Insertion/Deletion Metric-aware Regularizerscited by 0
- 2025TPK: Trustworthy Trajectory Prediction Integrating Prior Knowledge For Interpretability and Kinematic Feasibility10 May 2025
- 2024Linking Model Intervention to Causal Interpretation in Model Explanation21 Oct 2024
- 2024Causal Feature Selection for Responsible Machine Learning5 Feb 2024
- 2023Explanation-based Training with Differentiable Insertion/Deletion Metric-aware Regularizers19 Oct 2023
- 2023Learning Channel Importance for High Content Imaging with Interpretable Deep Input Channel Mixing31 Aug 2023
- 2023Neural Markov Jump Processes31 May 2023
- 2023Causal Discovery from Temporal Data: An Overview and New Perspectives17 Mar 2023
- 2021Training Deep Models to be Explained with Fewer Examples7 Dec 2021
- 2021Temporal-Spatial Causal Interpretations for Vision-Based Reinforcement Learning6 Dec 2021
- 2021Instance-wise Causal Feature Selection for Model Interpretation26 Apr 2021
- 2021Inductive Granger Causal Modeling for Multivariate Time Series10 Feb 2021
Similar, not cited
Close in wording, no citation either way
- 2026Neural Additive Experts: Context-Gated Experts for Controllable Model Additivity
- 2024Sparse Mixture-of-Experts for Compositional Generalization: Empirical Evidence and Theoretical Foundations of Optimal Sparsity
- 2026Neural Thickets: Diverse Task Experts Are Dense Around Pretrained Weights
- 2025Sparsity and Superposition in Mixture of Experts
- 2022Functional mixture-of-experts for classification
- 2023Improving Expert Specialization in Mixture of Experts
- 2025Mixture of Experts Provably Detect and Learn the Latent Cluster Structure in Gradient-Based Learning
- 2026Sparsely gated tiny linear experts