Tal Shnitzer, MIT: “Multi-modal and Temporal Data Analysis with Diffusion Operators”

Email: talsd@mit.edu

Position: Postdoctoral Researcher

Current Institution: MIT

Abstract: Multi-modal and Temporal Data Analysis with Diffusion Operators

Analysis and fusion of multi-modal data can be challenging especially when the data are high-dimensional and lack adequate models. We approach this problem from a manifold learning viewpoint in which the high-dimensional data is assumed to lie on some manifold of lower dimensions. Typically in this approach some data-driven operator that captures the geometrical properties of the data is constructed and a new representation of lower dimensions is obtained. Such methods can be especially useful when there is limited labeled data which is often the case in medical applications and have shown promising results in numerous applications. For example brain activity can be recorded using EEG which is commonly comprised of 32-64 electrodes positioned at different locations on the scalp. Due to the complex nature of such data obtaining a low-dimensional representation that captures its essence can be hard. These electrodes typically contain sensor specific noise and thus combining information from several electrodes can attenuate such interference leading to better representations of the general state and activity patterns of the brain. In our work we propose new ways of combining diffusion maps operators (a manifold learning technique) constructed from data captured by different modalities. Our method provides a theoretically justified spectral component analysis that separates the common components that are expressed similarly in the different modalities from those that are expressed differently. We demonstrate it on various applications including 3D shapes non-invasive fetal heart activity recovery from maternal measurements and remote sensing with hyperspectral and LiDAR images.


I am a postdoc at the Geometric Data Processing group at CSAIL MIT with Prof. Justin Solomon. My research interests include manifold learning data fusion biomedical signal processing and geometric data analysis. My research is supported by The Eric and Wendy Schmidt Postdoctoral Award for Women in Mathematical and Computing Sciences and by the Viterbi Award Technion. I received my B.Sc. degree summa cum laude in Electrical Engineering and in Biomedical Engineering from the Technion-IIT Israel in 2013 and my Ph.D. degree in Electrical Engineering from the Technion in 2020. My Ph.D. under the supervision of Prof. Ronen Talmon focused on developing new data-driven manifold learning methods for multi-modal and temporal data analysis.