Solving MRF Minimization by Mirror Descent
نویسندگان
چکیده
Markov Random Fields (MRF) minimization is a well-known problem in computer vision. We consider the augmented dual of the MRF minimization problem and develop a Mirror Descent algorithm based on weighted Entropy and Euclidean Projection. The augmented dual problem consists of maximizing a non-differentiable objective function subject to simplex and linear constraints. We analyze the convergence properties of the algorithm and sharpen its convergence rate. In addition, we also use the convergence analysis to identify an optimal stepsize strategy for weighted entropy projection and an adaptive stepsize strategy for weighted Euclidean projection. Experimental results on synthetic and vision problems demonstrate the effectiveness of our approach.
منابع مشابه
Mirror descent and nonlinear projected subgradient methods for convex optimization
The mirror descent algorithm (MDA) was introduced by Nemirovsky and Yudin for solving convex optimization problems. This method exhibits an e3ciency estimate that is mildly dependent in the decision variables dimension, and thus suitable for solving very large scale optimization problems. We present a new derivation and analysis of this algorithm. We show that the MDA can be viewed as a nonline...
متن کاملOn Stochastic Subgradient Mirror-Descent Algorithm with Weighted Averaging
This paper considers stochastic subgradient mirror-descent method for solving constrained convex minimization problems. In particular, a stochastic subgradient mirror-descent method with weighted iterate-averaging is investigated and its per-iterate convergence rate is analyzed. The novel part of the approach is in the choice of weights that are used to construct the averages. Through the use o...
متن کاملMirror-Descent-like Algorithms for Submodular Optimization
In this paper we develop a framework of submodular optimization algorithms in line with the mirror-descent style of algorithms for convex optimization. We use the fact that a submodular function has both a subdifferential and a superdifferential, which enables us to formulate algorithms for both submodular minimization and maximization. This reveals a unifying framework for a number of submodul...
متن کاملEfficient Methods for Stochastic Composite Optimization
This paper considers an important class of convex programming problems whose objective function Ψ is given by the summation of a smooth and non-smooth component. Further, it is assumed that the only information available for the numerical scheme to solve these problems is the subgradients of Ψ contaminated by stochastic noise. Our contribution mainly consists of the following aspects. Firstly, ...
متن کاملSparse Q-learning with Mirror Descent
This paper explores a new framework for reinforcement learning based on online convex optimization, in particular mirror descent and related algorithms. Mirror descent can be viewed as an enhanced gradient method, particularly suited to minimization of convex functions in highdimensional spaces. Unlike traditional gradient methods, mirror descent undertakes gradient updates of weights in both t...
متن کامل