Multilevel Methods for Sparse Representation of Topographical Data

With the onset of the data age, more and more of research is being carried out on construction of efficient algorithms for handling Big Data sets. The current work proposes a multiscale approach for intelligent prediction of missing data on a given DEM. The algorithm utilizes the Gaussian covariance kernel for studying the correlation between the available data points. Dimensionality reduction through parallel pivoted QR makes the approach scalable on large computing clusters. We have also studied the performance of the algorithm in terms of accuracy, scalability and convergence in order to validate the applicability of our approach. Use in generation of sparse representations for memory taxing datasets further establishes the efficiency of the multilevel analysis.