The inverse problem in microlensing