Abstract: We determine the initial temperature profile on a slab with adiabatic boundary condition, from a transient temperature distribution, obtained at a given time. This is an ill-posed inverse problem, where the initial condition has to be estimated. Two different neural networks have been applied to address the problem: backpropagation and radial basis functions (RBF). Both approaches use the whole temperature history mapping. In our simulations RBF presented better solutions, faster training, but higher noise sensitiveness, as compared to backpropagation.