Corpus ID: 212414796

What went wrong and when? Instance-wise Feature Importance for Time-series Models

  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},
  • Sana Tonekaboni, S. Joshi, +1 author Anna Goldenberg
  • Published 2020
  • Computer Science, Mathematics
  • ArXiv
  • 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
    5 Citations
    Benchmarking Deep Learning Interpretability in Time Series Predictions
    • PDF
    Learning to Evaluate Perception Models Using Planner-Centric Metrics
    • 5
    • PDF


    RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism
    • 451
    • PDF
    Attend and Diagnose: Clinical Time Series Analysis using Attention Models
    • 110
    • Highly Influential
    • PDF
    A Causality Based Feature Selection Approach for Multivariate Time Series Forecasting
    • 8
    Exploring Interpretable LSTM Neural Networks over Multi-Variable Data
    • 30
    • PDF
    An attention based deep learning model of clinical events in the intensive care unit
    • 33
    • PDF
    Explaining Therapy Predictions with Layer-Wise Relevance Propagation in Neural Networks
    • 27
    • PDF
    Clinical Intervention Prediction and Understanding using Deep Networks
    • 45
    • Highly Influential
    • PDF
    An interpretable LSTM neural network for autoregressive exogenous model
    • 15
    • PDF
    Explaining an increase in predicted risk for clinical alerts
    • 3
    • PDF
    TSViz: Demystification of Deep Learning Models for Time-Series Analysis
    • 17
    • PDF