Benchmarking Attention-Based Interpretability of Deep Learning in Multivariate Time Series Predictions
1
Department of Physics, Faculty of Science, University of Zagreb, Bijenička cesta 32, 10000 Zagreb, Croatia
2
Division of Electronics, Ruđer Bošković Institute, Bijenička cesta 54, 10000 Zagreb, Croatia
*
Authors to whom correspondence should be addressed.
Academic Editor: Sotiris Kotsiantis
Entropy 2021, 23(2), 143; https://doi.org/10.3390/e23020143
Received: 19 December 2020
/
Revised: 20 January 2021
/
Accepted: 21 January 2021
/
Published: 25 January 2021
(This article belongs to the Special Issue Human-Centric AI: The Symbiosis of Human and Artificial Intelligence)
The adaptation of deep learning models within safety-critical systems cannot rely only on good prediction performance but needs to provide interpretable and robust explanations for their decisions. When modeling complex sequences, attention mechanisms are regarded as the established approach to support deep neural networks with intrinsic interpretability. This paper focuses on the emerging trend of specifically designing diagnostic datasets for understanding the inner workings of attention mechanism based deep learning models for multivariate forecasting tasks. We design a novel benchmark of synthetically designed datasets with the transparent underlying generating process of multiple time series interactions with increasing complexity. The benchmark enables empirical evaluation of the performance of attention based deep neural networks in three different aspects: (i) prediction performance score, (ii) interpretability correctness, (iii) sensitivity analysis. Our analysis shows that although most models have satisfying and stable prediction performance results, they often fail to give correct interpretability. The only model with both a satisfying performance score and correct interpretability is IMV-LSTM, capturing both autocorrelations and crosscorrelations between multiple time series. Interestingly, while evaluating IMV-LSTM on simulated data from statistical and mechanistic models, the correctness of interpretability increases with more complex datasets.
View Full-Text
Keywords:
multivariate time series; attention mechanism; interpretability; synthetically designed datasets
▼
Show Figures
Figure 1
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited
-
Externally hosted supplementary file 1
Link: https://github.com/hc-xai/mts-interpretability-benchmark
Description: Code for generating synthetic multi-variate time series datasets of new benchmark for attention-based interpretability of deep learning in multivariate trime series forecasting tasks.
MDPI and ACS Style
Barić, D.; Fumić, P.; Horvatić, D.; Lipic, T. Benchmarking Attention-Based Interpretability of Deep Learning in Multivariate Time Series Predictions. Entropy 2021, 23, 143. https://doi.org/10.3390/e23020143
AMA Style
Barić D, Fumić P, Horvatić D, Lipic T. Benchmarking Attention-Based Interpretability of Deep Learning in Multivariate Time Series Predictions. Entropy. 2021; 23(2):143. https://doi.org/10.3390/e23020143
Chicago/Turabian StyleBarić, Domjan, Petar Fumić, Davor Horvatić, and Tomislav Lipic. 2021. "Benchmarking Attention-Based Interpretability of Deep Learning in Multivariate Time Series Predictions" Entropy 23, no. 2: 143. https://doi.org/10.3390/e23020143
Find Other Styles
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.
