What is Informer?

Informer is a Long Sequence Time-Series Forecasting (LSTF) model based on the Transformer architecture, designed to address the limitations of existing Transformer-based models when dealing with long input sequences. It introduces the ProbSparse self-attention mechanism to reduce computational complexity, the Self-Attention Distilling operation to highlight dominant attention patterns, and a generative style decoder to improve prediction speed. Informer is commonly used for forecasting tasks in various domains, including energy, finance, and healthcare, where accurate predictions over extended time horizons are crucial.