Time Series Model Attribution Visualizations as Explanations

TL;DR

This paper reviews attribution visualization techniques for time series models, discussing their advantages, disadvantages, and future opportunities, to improve interpretability of deep learning decisions.

Contribution

It provides a comprehensive collection and analysis of attribution visualizations for time series, highlighting their strengths and limitations, and suggests future directions.

Findings

Heatmaps are common but not always ideal for time series explanations.

Alternative visualization methods can complement heatmaps.

The paper identifies future opportunities for improving time series attributions.

Abstract

Attributions are a common local explanation technique for deep learning models on single samples as they are easily extractable and demonstrate the relevance of input values. In many cases, heatmaps visualize such attributions for samples, for instance, on images. However, heatmaps are not always the ideal visualization to explain certain model decisions for other data types. In this review, we focus on attribution visualizations for time series. We collect attribution heatmap visualizations and some alternatives, discuss the advantages as well as disadvantages and give a short position towards future opportunities for attributions and explanations for time series.