Sharing Information in Parallel Search with Search Space Partitioning

2013 ◽  
pp. 52-58 ◽  
Author(s):  
Davide Lanti ◽  
Norbert Manthey
Keyword(s):  
Search Space ◽  
Parallel Search ◽  
Space Partitioning

Search Space Partitioning to Enhance Outlier Rule Discovery

2002 ◽  
pp. 259-263
Author(s):  
Michał Okoniewski ◽  
Piotr Gawrysiak ◽  
Łukasz Gancarz
Keyword(s):  
Search Space ◽  
Rule Discovery ◽  
Space Partitioning

scholarly journals Parallel AND/OR Search for Marginal MAP

2020 ◽  
Vol 34 (06) ◽  
pp. 10226-10234
Author(s):  
Radu Marinescu ◽  
Akihiro Kishimoto ◽  
Adi Botea
Keyword(s):  
Graphical Models ◽  
State Of The Art ◽  
Search Algorithm ◽  
Search Space ◽  
Parallel Search ◽  
Sequential Search ◽  
Limited Memory ◽  
Mixed Inference ◽  
First Time ◽  
Inference Task

Marginal MAP is a difficult mixed inference task for graphical models. Existing state-of-the-art algorithms for solving exactly this task are based on either depth-first or best-first sequential search over an AND/OR search space. In this paper, we explore and evaluate for the first time the power of parallel search for exact Marginal MAP inference. We introduce a new parallel shared-memory recursive best-first AND/OR search algorithm that explores the search space in a best-first manner while operating with limited memory. Subsequently, we develop a complete parallel search scheme that only parallelizes the conditional likelihood computations. We also extend the proposed algorithms into depth-first parallel search schemes. Our experiments on difficult benchmarks demonstrate the effectiveness of the parallel search algorithms against current sequential methods for solving Marginal MAP exactly.

scholarly journals Parallelization algorithms for modeling ARM processes

2000 ◽  
Vol 13 (4) ◽  
pp. 393-410
Author(s):  
Benjamin Melamed ◽  
Santokh Singh
Keyword(s):  
Stochastic Processes ◽  
Objective Function ◽  
Empirical Distribution ◽  
Sample Path ◽  
Real Life ◽  
Search Space ◽  
Global Search ◽  
Communication Overhead ◽  
Space Partitioning ◽  
Nonlinear Objective Function

AutoRegressive Modular (ARM) processes are a new class of nonlinear stochastic processes, which can accurately model a large class of stochastic processes, by capturing the empirical distribution and autocorrelation function simultaneously. Given an empirical sample path, the ARM modeling procedure consists of two steps: a global search for locating the minima of a nonlinear objective function over a large parametric space, and a local optimization of optimal or near optimal models found in the first step. In particular, since the first task calls for the evaluation of the objective function at each vector of the search space, the global search is a time consuming procedure. To speed up the computations, parallelization of the global search can be effectively used by partitioning the search space among multiple processors, since the requisite communication overhead is negligible.This paper describes two space-partitioning methods, called Interleaving and Segmentation, respectively. The speedups resulting from these methods are compared for their performance in modeling real-life data.

Sign in / Sign up

Export Citation Format

Share Document