Michael Bronstein
1 min readAug 3, 2020

--

Thanks Arunprakash - it's hard to answer your question without actually seeing your results. One thing to have in mind is that eigenvectors of a matrix are defined up to scale, since in A*V=lambda*V we can multiply both sides by any constant. Typically the eigenvectors are L2-normalised, which is different from some of the definitions of the DFT/IDFT (signal processing books usually normalise IDFT by 1/n, whereas to assure it is orthonormal one has to use 1/sqrt(n) for both DFT and IDFT). This might be the source of your discrepancy.

--

--

Michael Bronstein
Michael Bronstein

Written by Michael Bronstein

DeepMind Professor of AI @Oxford. Serial startupper. ML for graphs, biochemistry, drug design, and animal communication.

No responses yet