Machine learning has revolutionized the way we do data science. In particular, deep neural networks have closed the performance gap between humans and machines in disciplines such as vision, image processing, audio processing, and natural language processing. Nonetheless, learning from unstructured high-dimensional empirical observations (for example gene measurements) is usually impossible for humans and may be difficult for machines. In particular, machine learning becomes extremely challenging since empirical measurements are often unlabeled, noisy, sparse, imbalanced, and high dimensional.
My main research goal is to develop automatic methods that would lead to novel scientific findings (in biology, physics, medicine, etc.). To achieve this goal I am currently developing deep learning methodologies that tackle the following tasks:
Feature Selection
Example paper - Feature selection using Stochastic Gates
Feature Extraction/ Manifold Learning
Example paper: LOCA: LOcal Conformal Autoencoder for standardized data coordinates
Multimodal Fusion
Example Paper: Multiview Diffusion Maps
Generative Models
Example Paper: Variational Diffusion Autoencoders with Random Walk Sampling
Biomedical Data Analysis
Example Paper: Alignment free identification of clones in B cell receptor repertoires
Signal Processing
Example Paper: Multiview Kernels for Low-Dimensional Modeling of Seismic Events