Optimization by Max-Propagation Using Kikuchi Approximations
نویسندگان
چکیده
In this paper we address the problem of using region-based approximations to find the optimal points of a given function. Our approach combines the use of Kikuchi approximations with the application of generalized belief propagation (GBP) using maximization instead of marginalization. The relationship between the fixed points of maximum GBP and the free energy is elucidated. A straightforward connection between the function to be optimized and the Kikuchi approximation (which holds only for maximum GBP, not for marginal GBP) is proven. Later, we show that maximum GBP can be combined with a dynamic programming algorithm to find the most probable configurations of a f graphical model. We then analyze the dynamics of the procedure proposed and show how its different steps can be manipulated to influence the search for optimal solutions.
منابع مشابه
Estimation and Marginalization Using the Kikuchi Approximation Methods
In this letter, we examine a general method of approximation, known as the Kikuchi approximation method, for finding the marginals of a product distribution, as well as the corresponding partition function. The Kikuchi approximation method defines a certain constrained optimization problem, called the Kikuchi problem, and treats its stationary points as approximations to the desired marginals. ...
متن کاملDecentralized Supply Chain Formation using Max-Sum Loopy Belief Propagation
Supply chain formation is the process by which a set of producers within a network determine the subset of these producers able to form a chain to supply goods to one or more consumers at the lowest cost. This problem has been tackled in a number of ways, including auctions, negotiations, and argumentation-based approaches. In this paper we show how this problem can be cast as an optimization o...
متن کاملA Max-Sum algorithm for training discrete neural networks
We present an efficient learning algorithm for the problem of training neural networks with discrete synapses, a well-known hard (NP-complete) discrete optimization problem. The algorithm is a variant of the so-called Max-Sum (MS) algorithm. In particular, we show how, for bounded integer weights with q distinct states and independent concave a priori distribution (e.g. l1 regularization), the ...
متن کاملBethe free energy, Kikuchi approximations, and belief propagation algorithms
This is an updated and expanded version of TR2000-26, but it is still in draft form. Belief propagation (BP) was only supposed to work for tree-like networks but works surprisingly well in many applications involving networks with loops, including turbo codes. However, there has been little understanding of the algorithm or the nature of the solutions it finds for general graphs. We show that B...
متن کامل(BP)2: Beyond pairwise Belief Propagation labeling by approximating Kikuchi free energies
Belief Propagation (BP) can be very useful and efficient for performing approximate inference on graphs. But when the graph is very highly connected with strong conflicting interactions, BP tends to fail to converge. Generalized Belief Propagation (GBP) provides more accurate solutions on such graphs, by approximating Kikuchi free energies, but the clusters required for the Kikuchi approximatio...
متن کامل