Abstract: Unfolding algorithms for neutron depth profiling:probability iteration, SVD solving least square, linear regularization(LR) and constrained linear reqularization(CLR) are studied. All algorithms are applied for both under-determined and over-determined equations, and results are compared and discussed. Due to iterative processes, probability iteration and CLR could not work well in the case that sources intensities change sharply. LR is unstable as unfolding range is chosen arbitrarily, which can be constrained by CLR. A practical spectrum of NDP experiment is unfolded by algorithms. LR could not restrain stochastic errors caused by statistical fluctuation, while other algorithms show good unfolding results and agree well with references.