search

thuml / Autoformer

About Code release for "Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting" (NeurIPS 2021), https://arxiv.org/abs/2106.13008
MIT License
1.99k stars 429 forks source link

top K #110

Closed oremnirv closed 1 year ago

oremnirv commented 1 year ago

Clarification question: In algorithm 2 in your paper, how do you go from the softmax weights having K dimensions to R variable with L dimensions?

Thank you

wuhaixu2016 commented 1 year ago

Hi, this process is quite clear in Code. Line 38-48 in https://github.com/thuml/Autoformer/blob/main/layers/AutoCorrelation.py We will adopt the K softmax weights to aggregate K delayed length-L time series.