Approximation of matrices by Gabor multipliers

We present a fast Fourier transform based algorithm which allows to obtain the best approximation of a given time-variant linear system by approximating its matrix (in the Hilbert-Schmidt sense) by a Gabor multiplier, using a given tight Gabor frame. We illustrate the approach by approximating the inverse of a slowly time variant filter.