Oded Regev

Oded Regev's lab uses interpretable machine learning techniques to derive insights into biological processes. Although they provide good predictions, typical machine learning models are 'black box' and as a result do not help advance our understanding of biology. Without a good understanding of the underlying processes, generalizing beyond the dataset (to other cell types, developmental states, diseases contexts, etc.) is nearly impossible. By designing our models with interpretability in mind, we are able to 'observe' molecular processes such as RNA processing or translation at the level of individual protein-nucleic acid interactions, resulting in novel models. My other work is in the areas of cryptography, quantum computation, and complexity theory. His main contribution there is to lattice-based cryptography, and specifically the Learning With Errors problem, forming the basis of post-quantum cryptography.