Event time:
Tuesday, November 18, 2014 - 11:00am to 12:00pm
Location:
AKW 000
Speaker:
Stefan Steinerberger
Speaker affiliation:
Yale University
Event description:
Spectral methods have proven to be a valuable tool for the task of mapping a high-dimensional set of points into low dimensions in a way that preserves or even extracts the relevant, underlying geometric structures. We describe a simple operation on Markov chains which can increase the effectivity of these methods and give many examples; we also prove that this method will automatically correct random errors with high probability.