Application of stochastic inverse theory to ionospheric tomography