Data Analyzing in Microfluidic Experiments with Machine Learning

Modern machine learning algorithms are implemented to help to extract the information from microfluidic experiments and to simplify the fluid-structure interaction simulation with reduced cost. In detail, this research involves 1) a multiplex application of a machine learning method for cell sorting, a label-free rare cell detection from images based on a deep learning model; 2) the discovery of the theory for building up a physics informed data-driven ”hybrid” model to solve steady Navier-stokes equations. The trained model can instantly generate the steady solutions of the Navier-Stokes equations with various boundary conditions.