Inherently Interpretable Time Series Classification via Multiple Instance Learning

TL;DR

This work proposes a new framework called MILLET: Multiple Instance Learning for Locally Explainable Time series classification, and applies it to existing deep learning TSC models and shows how they become inherently interpretable without compromising (and in some cases, even improving) predictive performance.

Abstract

Conventional Time Series Classification (TSC) methods are often black boxes that obscure inherent interpretation of their decision-making processes. In this work, we leverage Multiple Instance Learning (MIL) to overcome this issue, and propose a new framework called MILLET: Multiple Instance Learning for Locally Explainable Time series classification. We apply MILLET to existing deep learning TSC models and show how they become inherently interpretable without compromising (and in some cases, even improving) predictive performance. We evaluate MILLET on 85 UCR TSC datasets and also present a novel synthetic dataset that is specially designed to facilitate interpretability evaluation. On these datasets, we show MILLET produces sparse explanations quickly that are of higher quality than other well-known interpretability methods. To the best of our knowledge, our work with MILLET, which is available on GitHub (https://github.com/JAEarly/MILTimeSeriesClassification), is the first to develop general MIL methods for TSC and apply them to an extensive variety of domains

Figures (15)

• Figure 1: Conventional TSC techniques (left) usually only provide class-level predictive probabilities. In addition, our proposed method (MILLET, right) also highlights class-conditional discriminatory motifs that influence the predicted class. In the heatmap, green regions indicate support for the predicted class, red regions refute the predicted class, and darker regions are more influential.
• Figure 2: The five different MIL pooling methods used in this work. Each takes the same input: a bag of time point embeddings $\mathbf{Z_i} \in \mathbb{R}^{t \times d} = [\mathbf{z_i^1}, \mathbf{z_i^2}, \ldots, \mathbf{z_i^t}]$. While they all produce the same overall output (a time series prediction), they produce different interpretability outputs.
• Figure 3: An example for each class of our WebTraffic dataset. Signatures are injected in a single random window, with the exception of Class 1 (Spikes), which uses random individual time points.
• Figure 4: Interpretations for Conj.InceptionTime on our WebTraffic dataset. Top: Time series with the known discriminatory time points highlighted. Middle: Interpretability scores for each time point with respect to the target class. Bottom: Interpretability scores heatmap as in \ref{['fig:millet_overview']}.
• Figure 5: Critical difference diagram comparing ConjunctiveMILLET methods with SOTA.