ReCycle: Fast and Efficient Long Time Series Forecasting with Residual Cyclic Transformers

arXiv:2405.03429v1 Announce Type: new
Abstract: Transformers have recently gained prominence in long time series forecasting by elevating accuracies in a variety of use cases. Regrettably, in the race for better predictive performance the overhead of model architectures has grown onerous, leading to models with computational demand infeasible for most practical applications. To bridge the gap between high method complexity and realistic computational resources, we introduce the Residual Cyclic Transformer, ReCycle. ReCycle utilizes primary cycle compression to address the computational complexity …

arxiv cs.ai cs.lg forecasting residual series time series time series forecasting transformers type