Neural networks : the official journal of the International Neural Network SocietyForecastingHumansTime FactorsNeural NetworksComputer
MSA-LR: Enhancing multi-scale temporal dynamics in multivariate time series forecasting with low-rank self-attention.
Jie Sun, Zhilin Sun, Zhongshan Chen, Mengyang Dong, Xiaozheng Wang, Changwei Chen, Hao Zheng, Xiangjun Zhao
Published: 202610.1016/j.neunet.2025.108150
Abstract
Accurately forecasting multivariate time series requires effectively capturing intricate temporal dependencies across diverse scales. Existing deep learning models, while promising, often fall short in this regard. Recurrent architectures like LSTMs…
Preview only. Read the full abstract at the source