Fast Parallel and Adaptive Updates for Dual-Decomposition Solvers
Abstract
Dual-decomposition (DD) methods are quickly becoming important tools for estimating the minimum energy state of a graphical model. DD methods decompose a complex model into a collection of simpler subproblems that can be solved exactly (such as trees), that in combination provide upper and lower bounds on the exact solution. Subproblem choice can play a major role: larger subproblems tend to improve the bound more per iteration, while smaller subproblems enable highly parallel solvers and can benefit from re-using past solutions when there are few changes between iterations. We propose an algorithm that can balance many of these aspects to speed up convergence. Our method uses a cluster tree data structure that has been proposed for adaptive exact inference tasks, and we apply it in this paper to dual-decomposition approximate inference. This approach allows us to process large subproblems to improve the bounds at each iteration, while allowing a high degree of parallelizability and taking advantage of subproblems with sparse updates. For both synthetic inputs and a real-world stereo matching problem, we demonstrate that our algorithm is able to achieve significant improvement in convergence time.
Cite
Text
Sümer et al. "Fast Parallel and Adaptive Updates for Dual-Decomposition Solvers." AAAI Conference on Artificial Intelligence, 2011. doi:10.1609/AAAI.V25I1.8022Markdown
[Sümer et al. "Fast Parallel and Adaptive Updates for Dual-Decomposition Solvers." AAAI Conference on Artificial Intelligence, 2011.](https://mlanthology.org/aaai/2011/sumer2011aaai-fast/) doi:10.1609/AAAI.V25I1.8022BibTeX
@inproceedings{sumer2011aaai-fast,
title = {{Fast Parallel and Adaptive Updates for Dual-Decomposition Solvers}},
author = {Sümer, Özgür and Acar, Umut A. and Ihler, Alexander and Mettu, Ramgopal R.},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2011},
pages = {1076-1082},
doi = {10.1609/AAAI.V25I1.8022},
url = {https://mlanthology.org/aaai/2011/sumer2011aaai-fast/}
}