What went wrong and when? Instance-wise Feature Importance for Time-series Models
@article{Tonekaboni2020WhatWW, title={What went wrong and when? Instance-wise Feature Importance for Time-series Models}, author={Sana Tonekaboni and S. Joshi and D. Duvenaud and Anna Goldenberg}, journal={ArXiv}, year={2020}, volume={abs/2003.02821} }
Explanations of time series models are useful for high stakes applications like healthcare but have received little attention in machine learning literature. We propose FIT, a framework that evaluates the importance of observations for a multivariate time-series black-box model, by quantifying the shift in the predictive distribution over time. FIT defines the importance of an observation based on its contribution to the distributional shift under a KL-divergence that contrasts the predictive… CONTINUE READING
Figures, Tables, and Topics from this paper
5 Citations
Marginal Contribution Feature Importance - an Axiomatic Approach for The Natural Case
- Computer Science
- ArXiv
- 2020
- PDF
Benchmarking Deep Learning Interpretability in Time Series Predictions
- Computer Science, Mathematics
- NeurIPS
- 2020
- PDF
Learning to Evaluate Perception Models Using Planner-Centric Metrics
- Computer Science
- 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
- 2020
- 5
- PDF
References
SHOWING 1-10 OF 50 REFERENCES
RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism
- Computer Science, Mathematics
- NIPS
- 2016
- 451
- PDF
Attend and Diagnose: Clinical Time Series Analysis using Attention Models
- Psychology, Mathematics
- AAAI
- 2018
- 110
- Highly Influential
- PDF
A Causality Based Feature Selection Approach for Multivariate Time Series Forecasting
- Computer Science
- DBKDA 2017
- 2017
- 8
Exploring Interpretable LSTM Neural Networks over Multi-Variable Data
- Computer Science, Mathematics
- ICML
- 2019
- 30
- PDF
An attention based deep learning model of clinical events in the intensive care unit
- Medicine
- PloS one
- 2019
- 33
- PDF
Explaining Therapy Predictions with Layer-Wise Relevance Propagation in Neural Networks
- Computer Science
- 2018 IEEE International Conference on Healthcare Informatics (ICHI)
- 2018
- 27
- PDF
Clinical Intervention Prediction and Understanding using Deep Networks
- Mathematics, Computer Science
- ArXiv
- 2017
- 45
- Highly Influential
- PDF
An interpretable LSTM neural network for autoregressive exogenous model
- Computer Science, Mathematics
- ICLR
- 2018
- 15
- PDF
Explaining an increase in predicted risk for clinical alerts
- Computer Science, Mathematics
- CHIL
- 2020
- 3
- PDF
TSViz: Demystification of Deep Learning Models for Time-Series Analysis
- Computer Science, Mathematics
- IEEE Access
- 2019
- 17
- PDF