Stotts Steven A, Koch Robert A
Applied Research Laboratories, The University of Texas at Austin, P.O. Box 8029, Austin, Texas 78713-8029, USA.
J Acoust Soc Am. 2017 Aug;142(2):737. doi: 10.1121/1.4996456.
In this paper an approach is presented to estimate the constraint required to apply maximum entropy (ME) for statistical inference with underwater acoustic data from a single track segment. Previous algorithms for estimating the ME constraint require multiple source track segments to determine the constraint. The approach is relevant for addressing model mismatch effects, i.e., inaccuracies in parameter values determined from inversions because the propagation model does not account for all acoustic processes that contribute to the measured data. One effect of model mismatch is that the lowest cost inversion solution may be well outside a relatively well-known parameter value's uncertainty interval (prior), e.g., source speed from track reconstruction or towed source levels. The approach requires, for some particular parameter value, the ME constraint to produce an inferred uncertainty interval that encompasses the prior. Motivating this approach is the hypothesis that the proposed constraint determination procedure would produce a posterior probability density that accounts for the effect of model mismatch on inferred values of other inversion parameters for which the priors might be quite broad. Applications to both measured and simulated data are presented for model mismatch that produces minimum cost solutions either inside or outside some priors.
本文提出了一种方法，用于估计在对来自单个航迹段的水下声学数据进行统计推断时应用最大熵（ME）所需的约束。先前用于估计ME约束的算法需要多个源航迹段来确定该约束。该方法对于解决模型失配效应是相关的，即由于传播模型未考虑导致测量数据的所有声学过程，从反演确定的参数值存在不准确。模型失配的一个影响是，最低成本的反演解可能远在相对知名的参数值的不确定性区间（先验）之外，例如，来自航迹重建的源速度或拖曳源级。该方法对于某些特定参数值，要求ME约束产生一个包含先验的推断不确定性区间。提出这种方法的动机是这样一个假设，即所提出的约束确定过程将产生一个后验概率密度，该密度考虑了模型失配对其他反演参数推断值的影响，对于这些参数，先验可能相当宽泛。针对在某些先验之内或之外产生最低成本解的模型失配，给出了对实测数据和模拟数据的应用。
