Toward Transformer-Compatible Multivariate Time Series Learning via Visibility Graph-Based Structural Encoding

doi:10.1016/j.knosys.2025.114389

Toward Transformer-Compatible Multivariate Time Series Learning via Visibility Graph-Based Structural Encoding

2025-09-02

0

PRE

AI

DOI：10.1016/j.knosys.2025.114389

Original

Original request for help

Save

Abstract

En 中文

• Research highlight 1: Visibility Graph-guided Transformer focuses on learning crucial temporal patterns. • Research highlight 2: Sliding window visibility graph achieves linear complexity for large time series. • Research highlight 3: Consensus across variables improves understanding of global dependencies. • Research highlight 4: Embedding representations align well with the inherent features of the time series.

Toward Transformer-Compatible Multivariate Time Series Learning via Visibility Graph-Based Structural Encoding

Abstract

Journal

Organization