Skip to content


Track: Data Mining

Data Fusion Using Kernel Based Methods

Monday, April 12, 2:40-3:20pm EDT

Data fusion is a fundamental task in data-driven analysis. Datasets often contain several modalities of different types, each capturing one view of the problem. It is desired to form a compact, simple fused representation as a basis for analysis and explainability. Then, data mining tasks may be carried out in a reliable manner, while supported by a visual presentation of the complex data.

A kernel codes pairwise distances between items of similar type. By fusing two or more kernels that code different types of data, we overcome the heterogeneity challenge. Moreover, spectral decomposition of the fused kernels allow to visually plot and analyze the original data with a small number of latent factors. Specific applicative examples that benefit from this framework are discussed: performance monitoring for a time-evolving system, detecting mutual configurations of applied planning strategies and performances in small businesses, and modeling of a medical datasets.

Neta Rabin image

Neta Rabin

Neta Rabin

Senior Lecturer at the Department of Industrial Engineering, Tel-Aviv University

Neta Rabin received her B.Sc. degree in Mathematics and Computer Science, and an M.Sc degree in Computer Science, both from Tel-Aviv University. Neta completed her Ph.D in 2009 at the Department of Computer Science, Tel-Aviv University. Her Ph.D. thesis was focused on data mining in dynamically evolving systems via diffusion methodologies. Between 2009-2012, she was a Gibbs Assistant Professor at the Applied Mathematics Department at Yale University. Until 2019, Neta held a Senior Lecturer position in the Mathematics Department at Afeka, Tel-Aviv College of Engineering. Since 2019 she is a Senior Lecturer at the Department of Industrial Engineering, Tel-Aviv University. Her research is published in top-ranked journals and she serves as an Associate Editor for the Informs Data Science Journal.