Deep generative LDA

TL;DR

This paper introduces a deep generative LDA model based on discriminative normalization flow, demonstrating its superior ability to model complex data and improve low-dimensional representations over traditional LDA.

Contribution

It reinterprets DNF as a deep generative LDA, extending LDA's applicability to complex data distributions with improved modeling capabilities.

Findings

DNF outperforms traditional LDA in complex data modeling

Deep generative LDA provides better low-dimensional representations

Simulation and speaker recognition experiments validate the approach

Abstract

Linear discriminant analysis (LDA) is a popular tool for classification and dimension reduction. Limited by its linear form and the underlying Gaussian assumption, however, LDA is not applicable in situations where the data distribution is complex. Recently, we proposed a discriminative normalization flow (DNF) model. In this study, we reinterpret DNF as a deep generative LDA model, and study its properties in representing complex data. We conducted a simulation experiment and a speaker recognition experiment. The results show that DNF and its subspace version are much more powerful than the conventional LDA in modeling complex data and retrieving low-dimensional representations.