Blending ideas from discrete and continuous Markov processes with harmonic methods, diffusion maps are a tool for low-dimensional representation of data, seeking to reconstruct a low-dimensional submanifold upon which the clean data are thought to lie. Generally inexpensive to implement and fairly robust to noise, they are an important tool in modern machine learning.
Graduate Student Seminar: Introduction to Diffusion Maps
Feb 13 2018 - 12:30pm
Graduate Student Seminar
Fine Hall 214