Improving Transformer-based Multivariate Time Series Forecasting with Vector Field Embeddings and Sparse Attention

Proceedings