Answers are grounded in this paper and cite their sources.
Conventional EM optimization aims to use fewest possible fine model evaluations to increase the speed of optimization. In this work, we propose to use a large number of fine model evaluations to achieve an overall speedup. A large number of fine model evaluations allows us to build a surrogate model valid in a large neighborhood. In the proposed technique, these valid surrogate models are used to achieve large and effective optimization updates, thereby resulting in fewer iterations of the optimization process. Valid surrogate models uses many fine model evaluations which are realized in parallel using hybrid distributed shared memory computing platforms. Parallel computation of large number of fine model evaluations reduces the major computational time required for constructing a surrogate model. Furthermore, we exploit trust region algorithms to guarantee convergence and to re-define the fine model evaluation range in each iteration of the proposed optimization algorithm. The proposed technique aims to increase the speed of gradient based EM optimization when no coarse model (e.g., empirical or equivalent circuits) is available. Three typical examples are used to illustrate the proposed technique.
No takes yet. Share an insight, caveat, or question.
Gongal-Reddy et al. (2015) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: