B0710
Title: Sketched Gaussian model linear discriminant analysis via randomized Kaczmarz
Authors: Jocelyn Chi - UCLA (United States) [presenting]
Deanna Needell - UCLA (United States)
Abstract: Sketched linear discriminant analysis is presented, which is an iterative randomized approach to binary-class Gaussian model linear discriminant analysis (LDA) for very large data. We harness a least squares formulation and mobilize the stochastic gradient descent framework. Therefore, we obtain a randomized classifier with performance that is very comparable to that of full data LDA while requiring access to only one row of the training data at a time. We present convergence guarantees for the sketched predictions on new data within a fixed number of iterations. These guarantees account for both the Gaussian modeling assumptions on the training data and the algorithmic randomness in the sketching procedure. Finally, we demonstrate its performance with varying step-sizes and numbers of iterations. Our numerical experiments demonstrate that sketched LDA can offer a very viable alternative to full-data LDA when the data may be too large for full-data analysis.
