发表机构
Indian Institute of Technology Gandhinagar（印度理工学院甘地纳加尔分校）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究致密双星并合引力波搜索中随机模板库覆盖问题，提出用低差异序列构建随机模板库，可减少提议点数量，降低内存使用和簿记工作，为引力波搜索的随机模板库生成提供简单可扩展改进。
AI 中文摘要
匹配滤波仍然是检测致密双星并合产生的引力波最灵敏的方法。此类搜索的效率取决于离散模板库对潜在参数空间的覆盖程度。传统的几何、随机和混合放置方法在高维中可能导致覆盖不均匀和模板冗余。混合方法通常是这些方法中最有效的，而随机方法实现起来更简单，尤其是在参数空间度量难以计算时。在实践中，这两种方法都依赖均匀随机采样，这通常需要大量提议点才能实现足够的覆盖。我们发现，使用低差异序列构建的随机模板库在二维中所需提议点少27.5%，在三维中少12%，同时能达到相当的恢复率。最终模板数量仅略有变化（约1%），这与覆盖问题的度量体积约束一致。因此，低差异采样的主要好处是减少初始提议集的大小，从而降低内存使用并减少模板库生成期间的簿记工作。由于最终模板数量主要由目标参数空间的度量体积决定，实际加速比小于提议数量的减少。然而，低差异采样为当前和未来的引力波搜索提供了一种简单且可扩展的随机模板库生成改进方法。
英文摘要
Matched filtering remains the most sensitive method for detecting gravitational waves from compact binary coalescences. The efficiency of such searches depends on how well a discrete template bank covers the underlying parameter space. Conventional geometric, stochastic, and hybrid placement methods can lead to uneven coverage and redundant templates in higher dimensions. Hybrid methods are generally the most efficient among these, while stochastic methods are simpler to implement, particularly when the parameter-space metric is difficult to compute. In practice, both approaches rely on uniform random sampling, which often requires a large number of proposal points to achieve adequate coverage. We find that stochastic template banks constructed using low-discrepancy sequences achieve comparable recovery fractions while requiring 27.5\% fewer proposal points in two dimensions, about 12\% fewer in three dimensions and 17\% fewer in a four-dimensional eccentric equal-spin toy bank. The final template count is much less sensitive to the proposal sampling, consistent with the metric-volume constraints of the covering problem. We introduce a slab-wise coarse prefilter to accelerate candidate rejection for strongly anisotropic metrics. Applied to the 4D toy bank, it gives a total bank-construction speed-up of about $1.28\times$. The primary benefit of low-discrepancy sampling is therefore a reduction in the size of the initial proposal set, leading to lower memory usage and reduced bookkeeping during bank generation. These results show that low-discrepancy sampling provides a simple improvement to stochastic template-bank generation for current and future gravitational-wave searches.
Comments25 Pages, 6 Figures, 5 Table
