Event Date
Speaker: Zhenyu Liao (Postdoctoral Scholar, Statistics, UC Berkeley)
Title: "Performance-complexity trade-off in large dimensional spectral clustering"
Abstract: The big data revolution comes along with the challenging need to parse, mine, compress large amount of large dimensional data. Many modern machine learning algorithms (including state-of-the-art deep neural networks) are designed to work with compressed, quantized, or even binarized data so that they can run on low-power IoT devices.
In this talk, we will focus on the theoretical analysis of spectral clustering method that aims to nd possible clusters from a given data matrix in an unsupervised manner, by exploring the informative eigenstructure (e.g., the dominant eigenvector) of the data matrix. Random matrix analysis reveals the surprising fact that very little change occurs in the informative eigenstructure
even under drastic sparsi cation and/or quantization, and consequently that very little downstream performance loss occurs with very aggressively uniformed and non-uniformed, sparsi ed and/or quantized spectral clustering. The present analysis is based on a spiked model-type analysis of nonlinear random matrices and may be of independent research interest. We expect that our analysis opens the door to improved analysis of computationally efficient methods for large dimensional machine learning and neural network models more generally.
About the speaker: Zhenyu Liao is currently a postdoctoral researcher at the University of California, Berkeley, Department of Statistics (hosted by Michael Mahoney). He received his Ph.D. from Centrale Supélec, University Paris-Saclay in 2019. His research interests are broadly in machine learning, signal processing, random matrix theory, and high-dimensional statistics.
Seminar Date/Time: Thursday January 21, 4:10pm
This seminar will be delivered remotely via Zoom. To access the Zoom meeting for this seminar, please contact the instructor Xiucai Ding (xcading@ucdavis.edu) or Pete Scully (pscully@ucdavis.edu) for the meeting ID and password, stating your affiliation.