Inference on Graphs: Iterative Maximization of Pseudo Log-map Functions
نویسندگان
چکیده
In this paper, we formalize and study the properties of a new kind of iterative estimation algorithm which has recently appeared in the literature in e.g. [1, 2, 3]. We refer to this algorithm as the ”iterative pseudo log-MAP function maximization (IPLFM) algorithm. We give a definition of the pseudo log-MAP function (PLF) in terms of regions of a factor graph and prove some of its properties. In particular, we provide a correspondence between the zeroth, first and second order behaviors of the PLF and the (minimum) Bethe free energy associated to the considered factor graph. Based on these properties, we prove some results pertaining to the fixed points and the local convergence of the IPLFM algorithm. In particular, we relate the fixed points of the IPLFM-algorithm to the stationary points of the Bethe free energy and to the fixed points of the EM algorithm. Moreover, we provide necessary and sufficient conditions for the local convergence of the IPLFM algorithm. Key-words: MAP estimation, iterative methods, convergence of numerical methods, EM algorithm.
منابع مشابه
Belief Propagation, Mean-field, and Bethe approximations
This chapter describes methods for estimating the marginals and maximum a posteriori (MAP) estimates of probability distributions defined over graphs by approximate methods including Mean Field Theory (MFT), variational methods, and belief propagation. These methods typically formulate this problem in terms of minimizing a free energy function of pseudomarginals. They differ by the design of th...
متن کاملReduce and Re-Lift: Bootstrapped Lifted Likelihood Maximization for MAP
By handling whole sets of indistinguishable objects together, lifted belief propagation approaches have rendered large, previously intractable, probabilistic inference problems quickly solvable. In this paper, we show that Kumar and Zilberstein’s likelihood maximization (LM) approach to MAP inference is liftable, too, and actually provides additional structure for optimization. Specifically, it...
متن کاملTree consistency and bounds on the performance of the max-product algorithm and its generalizations
Finding the maximum a posteriori (MAP) assignment of a discrete-state distribution specified by a graphical model requires solving an integer program. The max-product algorithm, also known as the max-plus or min-sum algorithm, is an iterative method for (approximately) solving such a problem on graphs with cycles. We provide a novel perspective on the algorithm, which is based on the idea of re...
متن کاملNear-Optimal MAP Inference for Determinantal Point Processes
Determinantal point processes (DPPs) have recently been proposed as computationally efficient probabilistic models of diverse sets for a variety of applications, including document summarization, image search, and pose estimation. Many DPP inference operations, including normalization and sampling, are tractable; however, finding the most likely configuration (MAP), which is often required in p...
متن کاملImplicit Flow Maximization by Iterative Squaring
Application areas like logic design and network analysis produce large graphs G = (V,E) on which traditional algorithms, which work on adjacency list representations, are not practicable anymore. These large graphs often contain regular structures that enable compact implicit representations by decision diagrams like OBDDs [1], [2], [3]. To solve problems on such implicitly given graphs, specia...
متن کامل