forieux

openpgp4fpr:17218F250C9AB128A04B29591131BB2904F15681

forieux

This Gist compares JAX, Numpy and Scipy. The comparison is about the computation time (CPU only) of gradient of the criterion

$J(x) = |y - H x|^2$

where $H = F^H Λ F W$ is a linear operator with $F$ the Discrete Fourier transform, $Λ$ a complex diagonal matrix (then $FᴴΛF$ is circulant convolution), $W$ a diagonal matrix, $x$ and $y$ are images.

Gradients

forieux

### Keybase proof

I hereby claim:

  * I am forieux on github.
  * I am forieux (https://keybase.io/forieux) on keybase.
  * I have a public key ASDPHZj2BEzAuCfoQRqWmnKxdU8eSLJuU3VVXGubNTgzawo

To claim this, I am signing this object: