Series decomposition block

AutoCorrelation Mechanism with the following two phases: (1) period-based dependencies discovery (2) time delay aggregation This bloc

Special designed layernorm for the seasonal part

Autoformer encoder

Autoformer decoder layer with the progressive decomposition architecture

Autoformer encoder

Autoformer encoder layer with the progressive decomposition architecture

Moving average block to highlight the trend of time series

De-stationary Attention

Day of month encoded as value between [-0.5, 0.5]

Hour of day encoded as value between [-0.5, 0.5]

Day of year encoded as value between [-0.5, 0.5]

Hour of day encoded as value between [-0.5, 0.5]

Minute of hour encoded as value between [-0.5, 0.5]

Paper link: https://arxiv.org/pdf/2205.13504.pdf

Autoformer is the first method to achieve the series-wise connection, with inherent O(LlogL) complexity Paper link: https://openreview.ne

Month of year encoded as value between [-0.5, 0.5]

Minute of hour encoded as value between [-0.5, 0.5]

Week of year encoded as value between [-0.5, 0.5]

dot.notation access to dictionary attributes

Multiple Series decomposition block from FEDformer