DC Field | Value | Language |
---|---|---|
dc.contributor.author | Yoo, Jaemin | ko |
dc.contributor.author | Kang, U | ko |
dc.date.accessioned | 2023-09-05T01:02:43Z | - |
dc.date.available | 2023-09-05T01:02:43Z | - |
dc.date.created | 2023-09-04 | - |
dc.date.issued | 2021-04-30 | - |
dc.identifier.citation | SIAM International Conference on Data Mining, pp.531 - 539 | - |
dc.identifier.uri | http://hdl.handle.net/10203/312198 | - |
dc.description.abstract | Given a multivariate time series, how can we forecast all of its variables efficiently and accurately? The multivariate forecasting, which is to predict the future observations of a multivariate time series, is a fundamental problem closely related to many real-world applications. However, previous multivariate models suffer from large model sizes due to the inefficiency of capturing complex intra-variable patterns and inter-variable correlations, resulting in poor accuracy. In this work, we propose AttnAR (attention-based autoregression), a novel approach for general multivariate forecasting which maximizes its model efficiency via separable structure. AttnAR first extracts variable-wise patterns by a mixed convolution extractor that efficiently combines deep convolution layers and shallow dense layers. Then, AttnAR aggregates the patterns by learning time-invariant attention maps between the target variables. AttnAR accomplishes the state-of-the-art forecasting accuracy in four datasets with up to 117.3 times fewer parameters than the best competitors. | - |
dc.language | English | - |
dc.publisher | Society for Industrial and Applied Mathematics | - |
dc.title | Attention-Based Autoregression for Accurate and Efficient Multivariate Time Series Forecasting | - |
dc.type | Conference | - |
dc.type.rims | CONF | - |
dc.citation.beginningpage | 531 | - |
dc.citation.endingpage | 539 | - |
dc.citation.publicationname | SIAM International Conference on Data Mining | - |
dc.identifier.conferencecountry | US | - |
dc.identifier.conferencelocation | Virtual | - |
dc.identifier.doi | 10.1137/1.9781611976700.60 | - |
dc.contributor.localauthor | Yoo, Jaemin | - |
dc.contributor.nonIdAuthor | Kang, U | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.