Using Persistent Homology to understand High Dimensional Point Cloud DatasetsStudent-Run Research Seminar
|Speaker:||Greg Depaul, UC Davis|
|Start time:||Thu, May 19 2022, 12:10PM|
Current methods of correlation rely heavily on assumptions of linearity or monotonicity. We suggest leveraging the nonparametric power of Topological Data Analysis (TDA) in order to solve this problem. Specifically, we have developed a robust statistic that subdivides a point cloud data set into regions and uses TDA in order to conclude that a point cloud would produce a learnable function by a machine learning model.
Much of this talk will be entry level friendly, with a significant portion devoted to bringing up to speed anyone who would like to possibly research TDA in the future.