TIMELIGHT: EXPLAINABILITY IN TIME SERIES

Abstract

The PI introduces TIMELIGHT, an end-to-end System for explainability in time series; investigating how to build an effective and scalable system for explainability of time-series models. The main focus is on explainability in blackbox models of time series. There are four critical aspects one must address to develop a successful solution. They include: (1) Explaining Predictions in Time Series, (2) Explaining High-Dimensional Predictions in Time Series, (3) Interpretably Closing the Loop through Explanations and (4) End-to-End System for Explainability in Time Series.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 12, 2021
Source ID
FA95502010427

Entities

People

  • Emily B. Fox

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of Washington

Tags

Fields of Study

  • Computer science

Readers

  • Neural Network Machine Learning.