Multiscale Sparse Methods for Data Reduction and Modeling

In this project we developed a multiscale data reduction approach, that besides reducing the size of large datasets by exploiting the inherent redundancy, also provides an efficient sparse model that can be used for a variety of downstream learning tasks using regression and classsification as a building block.

Relevant publications for this project

  1. Shekhar P., Babu M., and Patra A. Hierarchical regularization networks for sparsification based learning on noisy datasets. Foundations of Data Science. 2023; doi: 10.3934/fods.2023009 Link arXiv Code.

  2. Shekhar P. and Patra A., A Forward Backward Greedy Approach for Sparse Multiscale Learning. Computer Methods in Applied Mechanics and Engineering. 2022; 400: 115420 Link arXiv Code.

  3. Shekhar P. and Patra A., Hierarchical approximations for data reduction and learning at multiple scales. Foundations of Data Science. 2020;2(2):123-154 Link arXiv Code.