Glinear

GLinear is a simplified time series forecasting model that integrates a non-linear GeLU transformation layer and a Reversible Instance Normalization (RevIN) component. This model captures complex patterns through the non-linear transformation layer while using RevIN to enhance training efficiency and stability. Due to its simple architecture, GLinear trains much faster than other Transformer-based predictors, while maintaining performance comparable to state-of-the-art forecasting models, making it highly valuable for practical applications.