ML Anthology

Partial Correlation-Based Attention for Multivariate Time Series Forecasting

Won Kyung Lee
AAAI 2020 pp. 13720-13721
doi:10.1609/AAAI.V34I10.7132 /aaai/2020/lee2020aaai-partial/

Abstract

A multivariate time-series forecasting has great potentials in various domains. However, it is challenging to find dependency structure among the time-series variables and appropriate time-lags for each variable, which change dynamically over time. In this study, I suggest partial correlation-based attention mechanism which overcomes the shortcomings of existing pair-wise comparisons-based attention mechanisms. Moreover, I propose data-driven series-wise multi-resolution convolutional layers to represent the input time-series data for domain agnostic learning.

PDF AAAI Semantic Scholar

Cite

Text

Lee. "Partial Correlation-Based Attention for Multivariate Time Series Forecasting." AAAI Conference on Artificial Intelligence, 2020. doi:10.1609/AAAI.V34I10.7132

Markdown

[Lee. "Partial Correlation-Based Attention for Multivariate Time Series Forecasting." AAAI Conference on Artificial Intelligence, 2020.](https://mlanthology.org/aaai/2020/lee2020aaai-partial/) doi:10.1609/AAAI.V34I10.7132

BibTeX

@inproceedings{lee2020aaai-partial,
  title     = {{Partial Correlation-Based Attention for Multivariate Time Series Forecasting}},
  author    = {Lee, Won Kyung},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2020},
  pages     = {13720-13721},
  doi       = {10.1609/AAAI.V34I10.7132},
  url       = {https://mlanthology.org/aaai/2020/lee2020aaai-partial/}
}