Parallel, Asychronous Global Optimization Techniques for Medium and Large Inversion Problems

用于地震射线追踪等问题的计算软件 三维复杂地质介质旅行时反演 通常只提供局部解决方案，即，局部最优， 而不是全球性的解决方案，因为“颠簸”的反应 待探索的表面。 这项研究将开发算法 可以使用粗粒度并行从局部最优值中逃脱 技术. 全局优化技术将与局部优化技术相结合。 优化技术解决多目标优化和 大反演问题。 这些发展的关键 算法是粗粒度并行的有效利用 异步运行。

Computing software for problems like seismic ray tracing and travel time inversion in three-dimensional complex geologic media often provides only local solutions, i.e., local optima, rather than global solutions because of the "bumpiness" of the response surface to be explored. This research will develop algorithms that can escape from local optima using coarse grain parallel techniques. Global optimization techniques will be combined with local optimization techniques to solve multiobjective optimization and large inversion problems. The key to the development of these algorithms is the efficient use of coarse grain parallelism operating asynchronously.