Elia Brenner
Pronouns: he/him
Research Mentor(s): Shuyang Cheng
Research Mentor School/College/Department: DCMB / Medicine
Program: UROP
Session: Session 6 (3:40pm – 4:30pm)
Authors: Elia Brenner, Shuyang Cheng
Abstract
The rapid growth in high volume biomedical datasets, which are collected from multiple modalities of heterogeneous nature and contain missing values, poses problems for the current methods of matrix-oriented data representation and many analysis algorithms. One promising solution for the computational challenges of analyzing such datasets is offered by the approach of tensor-oriented algorithms for data representation, decomposition and completion. A tensor is a multidimensional array where each modality spans one dimension. One common approach for analyzing data represented as tensors is to use various tensor decomposition methods which could significantly reduce the time complexity and computational resources required, by utilizing the multimodal structure in the data. Additionally, tensor decomposition also leads to various tensor completion methods which could be used to address the issue of missing data and develop new insights into the underlying structure of the data. One prominent example is the application of matrix and tensor methods to the combinatorial property of graphs, which arise from heterogeneous data that cannot be efficiently represented in Euclidean space. The aim of this research is to see if tensor based methods are an effective tool leading to real world medical applications.
