Ensemble Data Fitting

In regions with sparse bathymetry, data learning algorithms have shown skill in recognizing dominant features such as seamounts and ridges. The structure of these features provides a means to impute data values to increase the resolution. When two different types of classifiers identify the same acreage – we have two possible interpretations of the sparse data. In this paper we construct an ensemble data fitting method, designed for sparse Bathymetric acreage that arbitrates between two competing nominal data categories. Each categorical data type leads to different data imputation interpretations. From these two interpretations, we construct an ensemble regression to minimize a weighted average of the two categorical interpretations. We demonstrate the method using an idealized Bathymetric data set from which two interpretations are possible.