Abstract:Abstract Solving vision problems often entails searching a solution space for optimal state(s) that has maximum,Bayesian posterior probability or minimum,energy. When the volume of the space is huge, exhaustive search becomes infeasible. Generic stochastic search (e.g. Markov chain Monte Carlo) could be even worse than exhaustive search as it may visit a state repeatedly. To expedite the Markov chain search, one may use heuristics as proposal probability to guide the search in promising portions of the space. Empirically the recent data-driven Markov chain Monte Carlo (DDMCMC) scheme[14, 12, 2], achieves fast search in a number of vision tasks,intuitively justifled by two observations (i). The posterior probabilities in vision tasks often have very low entropy and thus are narrowly focused on a tiny portion of the state space (ii). The proposal probability computed in bottom-up methods can approximate the posterior well. In this paper, we study an independent Metropolis sampler which is often used in designing components of complex MCMC algorithms. We obtain an analytic formula for the expected time to flrst hit a certain state (\flrst hitting-time"), as well as very tight lower and upper bounds which depend on the total variation between the target posterior and the heuristic probabilities. These results show, though in humble cases, that one can indeed reach optimal solutions in very few steps with good proposal probabilities regardless of the size of the original search space. This result is difierent from previous analysis on the Markov chain convergence rate which is bounded by the second largest eigen-value (modulus) and often corresponds to the worst case in the entire search space. In comparison, our analysis bears more relevance to the optimization tasks in vision. Keywords: Markov Chain Monte Carlo, First Hitting Time, Convergence Rate, Inde-

The Computation of the Mean First Passage Times for Markov Chains

Higher Order Ergodic Markov Chains and First Passage Times

Mean hitting times of quantum Markov chains in terms of generalized inverses

Computing first passage times for Markov-modulated fluid models using numerical PDE problem solvers

A Computational Approach to First Passage Problems of Reflected Hyper-Exponential Jump Diffusion Processes

Efficient computation of first passage times in Kou's jump-diffusion model

Solving Markov Decision Processes with Reachability Characterization from Mean First Passage Times

Exact analytic expressions for discrete first-passage time probability distributions in Markov networks

Accurately Computing Expected Visiting Times and Stationary Distributions in Markov Chains

Efficient Computation of Mean Truncated Hitting Times on Very Large Graphs

Collective marks and first passage times

A NEW ALGORITHM FOR COMPUTING THE RATE MATRIX OF GI/M/1 TYPE MARKOV CHAINS

First Passage Times for Markov Renewal Processes and Applications

Waiting times and stopping probabilities for patterns in Markov chains

How Do Heuristics Expedite Markov Chain Search

Mean First Passage Times for Transport Equations

A First-Order Gradient Approach for the Connectivity Analysis of Markov Chains

Empirical and Instance-Dependent Estimation of Markov Chain and Mixing Time

Title How Do Heuristics Expedite Markov Chain Search ? Hitting-time Analysis of the Independence Metropolis Sampler Permalink

Fast Computation of Stationary Joint Probability Distribution of Sparse Markov Chains

The Distribution of Mixing Times in Markov Chains