search

aikunyi / FourierGNN

Official implementation of the paper "FourierGNN: Rethinking Multivariate Time Series Forecasting from a Pure Graph Perspective"
MIT License
139 stars 14 forks source link

Convolution theorem #4

Open mounchiliu opened 7 months ago

mounchiliu commented 7 months ago

Hi there, The code performs conduct DFT on NT dimension and then conducts matrix multiplication on d dimension. Why this kind of matrix multiplication in d dimension is equivalent to matrix multiplication AXW in the time domain?

aikunyi commented 7 months ago

The graph convolution operation AXW is also conducted on d dimension