Electron Flow Matching for Generative Reaction Mechanism Prediction | LeMaterial Reading Group скачать в хорошем качестве

Electron Flow Matching for Generative Reaction Mechanism Prediction | LeMaterial Reading Group

Paper Link: https://www.nature.com/articles/s4158... Most recent ML models in reaction prediction often fail to preserve mass conservation and are unable to generate/recover potential mechanistic pathways towards products. Thus, we propose FlowER (Flow Matching for Electron Redistribution), which models chemical reactions as a generative process of electron redistribution and uniquely combines 3 principles in an elegant fashion: mass conservation, mechanistic insights, and generative modelling to better align ML models with physical reality while remaining inherently explainable. Mun Hong is currently a first-year computer science graduate student at Duke University, advised by Rohit Singh and Alexander Tong. Previously, he was a software engineer/research assistant at MIT, working on AI for synthesis planning under the supervision of Connor W. Coley. LeMaterial Reading Group is a recurring gathering where we discuss recent papers at the intersection of AI, chemistry and materials science