Title: Mapping time-dependent magnetic topologies of active stars

Abstract: Throughout the last decades, Zeeman-Doppler Imaging (ZDI) has been intensively used to reconstruct large-scale magnetic topologies of active stars from time series of circularly polarized (Stokes $V$) profiles. ZDI being based on the assumption that the topology to be reconstructed is constant with time (apart from being sheared by differential rotation), it fails at describing stellar magnetic fields that evolve on timescales similar to the observing period. We present a new approach, called TIMeS (for Time-dependent Imaging of Magnetic Stars), to derive the time-dependent large-scale magnetic topologies of active stars, from time series of high-resolution Stokes $V$ spectra. This new method uses the combined concepts of sparse approximation and Gaussian process regression to derive the simplest time-dependent magnetic topology consistent with the data. Assuming a linear relation between the Stokes $V$ data and the reconstructed magnetic image, TIMeS is currently applicable to cases in which the magnetic field is not too strong (with an upper limit depending on $v\sin{i}$). We applied TIMeS to several simulated data sets to investigate its ability to retrieve the poloidal and toroidal components of large-scale magnetic topologies. We find that the proposed method works best in conditions similar to those needed for ZDI, reconstructing reliable topologies with minor discrepancies at very low latitudes whose contribution to the data is small. We however note that TIMeS can fail at reconstructing the input topology when the field evolves on a timescale much shorter than the stellar rotation cycle