Faculty Organizer: Prof. Corina Calinescu
Thursday, September 15, 2022, 1-2 pm in room N702
Speaker: Prof. Francis Patricia Medina (City Tech)
Title: How Mathematics can improve Machine Learning in LiDAR point clouds
Abstract: The main purpose of this talk is to show how mathematical tools and ideas can improve machine learning frameworks in datasets such as 3D LiDAR point clouds. The LiDAR point clouds contain measurements of complicated natural scenes and the classes are comprised of natural features such as ”foliage” and ”bare ground”. We present a preliminary comparison study for the classification of 3D point cloud LiDAR data that includes several types of feature engineering. We also include experiments with several dimension reduction strategies, ranging from Principal Component Analysis (PCA) to neural network-based auto-encoders, and demonstrate how they affect classification performance in LiDAR point clouds. As time allows, we will also give a general idea of multi-manifold approximating a point cloud and the computation of product coefficients (coming from measure theory) on a point cloud in order to showcase more mathematical ideas enriching machine learning applied to the sciences. Hopefully, I will be able to tell you how I am planning on connecting my machine learning research to my previous research in mathematical models involving methane.