1: \begin{abstract}

2: The estimation of a random vector with independent components passed through

3: a linear transform followed by a componentwise (possibly nonlinear) output map

4: arises in a range of applications. Approximate message passing (AMP) methods,

5: based on Gaussian approximations

6: of loopy belief propagation, have recently attracted considerable attention

7: for such  problems. For large random transforms, these methods

8: exhibit fast convergence and admit precise analytic characterizations with

9: testable conditions for optimality, even for certain non-convex problem instances.

10: However, the behavior of AMP under general transforms is not fully understood.  In this paper,

11: we consider the generalized AMP (GAMP) algorithm and relate the method to

12:  more common optimization techniques.  This analysis

13: enables a precise characterization of the GAMP algorithm fixed-points that applies to

14: arbitrary transforms.  In particular, we show that the fixed points

15: of the so-called max-sum GAMP algorithm for MAP estimation

16: are critical points of a constrained maximization of the posterior density.

17: The fixed-points of the sum-product GAMP algorithm for estimation of the posterior marginals

18: can be interpreted as critical points of a certain mean-field variational optimization.

19: \end{abstract}