Learning a Discriminant Latent Space with Neural Discriminant Analysis

TL;DR

This paper introduces Neural Discriminant Analysis (NDA), an optimization for deep neural networks that enhances discriminative features, leading to improved performance across classification, semi-supervised learning, and out-of-distribution detection tasks.

Contribution

The paper proposes NDA, a novel optimization inspired by LDA, for deep networks to better separate classes by reducing intra-class variance and increasing inter-class distances.

Findings

NDA improves classification accuracy across multiple datasets.

NDA surpasses state-of-the-art methods in four different tasks.

NDA enhances feature discriminability in deep neural networks.

Abstract

Discriminative features play an important role in image and object classification and also in other fields of research such as semi-supervised learning, fine-grained classification, out of distribution detection. Inspired by Linear Discriminant Analysis (LDA), we propose an optimization called Neural Discriminant Analysis (NDA) for Deep Convolutional Neural Networks (DCNNs). NDA transforms deep features to become more discriminative and, therefore, improves the performances in various tasks. Our proposed optimization has two primary goals for inter- and intra-class variances. The first one is to minimize variances within each individual class. The second goal is to maximize pairwise distances between features coming from different classes. We evaluate our NDA optimization in different research fields: general supervised classification, fine-grained classification, semi-supervised learning, and out of distribution detection. We achieve performance improvements in all the fields compared to baseline methods that do not use NDA. Besides, using NDA, we also surpass the state of the art on the four tasks on various testing datasets.