hrvatski jezikClear Cookie - decide language by browser settings

Benchmarking attention-based interpretability of deep learning in multivariate time series predictions

Barić, Domjan; Fumić, Petar; Horvatić, Davor; Lipić, Tomislav (2021) Benchmarking attention-based interpretability of deep learning in multivariate time series predictions. Entropy (Basel. Online), 23 (2). ISSN 1099-4300

PDF - Published Version - article
Available under License Creative Commons Attribution.

Download (729kB) | Preview


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.

Item Type: Article
Uncontrolled Keywords: multivariate time series ; attention mechanism ; interpretability ; synthetically designed datasets
Subjects: NATURAL SCIENCES > Physics
TECHNICAL SCIENCES > Interdisciplinary Technical Sciences
Divisions: Division of Electronics
Project titleProject leaderProject codeProject type
Provedba vrhunskih istraživanja u sklopu Znanstvenog centra izvrsnosti za kvantne i kompleksne sustave te reprezentacije Liejevih algebriBuljan, Hrvoje; Pandžić, PavleKK.
Napredne metode i tehnologije u znanosti o podatcima i kooperativnim sustavimaUNSPECIFIEDUNSPECIFIEDEK
Depositing User: Tomislav Lipić
Date Deposited: 21 Jan 2022 15:10
DOI: 10.3390/e23020143

Actions (login required)

View Item View Item


Downloads per month over past year

Increase Font
Decrease Font
Dyslexic Font