“…We assumed the forward model to operate on an unstructured piecewise polynomial finite element basis representation, while the solution of the inverse problem is represented by a regular grid of basis functions, be it a piecewise constant or polynomial pixel or voxel representation, or a blob basis representation with radially symmetric basis functions arranged in a regular grid. Adaptive approaches to the numerical solution of inverse problems, such as those described in the introduction, that could potentially benefit from the basis mapping algorithms derived in this paper include electromagnetic imaging, where the forward model is given by Maxwell's equations [23,24], diffuse optical tomography [4,25] and fluorescence tomography [26,27], electrical impedance tomography [28], as well as in adaptive inverse electrocardiography [29] and electroencephalography problems [30]. Applications can also be found in geophysics, using adaptive methods for seismic tomography [31], local earthquake tomography [32] and magnetic susceptibility mapping [33].…”