approximate message passing (AMP) is a low-cost iterative
parameter-estimation technique for certain high-dimensional linear systems with
non-Gaussian distributions. AMP only applies to independent identically
distributed (IID) transform matrices, but may become unreliable (e.g., perfo