Ensemble Soft-Margin Softmax Loss for Image Classification

TL;DR

This paper introduces EM-Softmax, a novel loss function that enhances feature discriminability and classifier strength in CNNs for image classification by incorporating soft-margin and ensemble techniques.

Contribution

It proposes a combined framework that improves feature discrimination and classifier robustness by integrating soft-margin softmax and ensemble methods using HSIC.

Findings

Outperforms baseline softmax loss on benchmark datasets.

Achieves higher feature discriminability and classification accuracy.

Demonstrates the effectiveness of ensemble classifiers with HSIC diversity.

Abstract

Softmax loss is arguably one of the most popular losses to train CNN models for image classification. However, recent works have exposed its limitation on feature discriminability. This paper casts a new viewpoint on the weakness of softmax loss. On the one hand, the CNN features learned using the softmax loss are often inadequately discriminative. We hence introduce a soft-margin softmax function to explicitly encourage the discrimination between different classes. On the other hand, the learned classifier of softmax loss is weak. We propose to assemble multiple these weak classifiers to a strong one, inspired by the recognition that the diversity among weak classifiers is critical to a good ensemble. To achieve the diversity, we adopt the Hilbert-Schmidt Independence Criterion (HSIC). Considering these two aspects in one framework, we design a novel loss, named as Ensemble soft-Margin Softmax (EM-Softmax). Extensive experiments on benchmark datasets are conducted to show the superiority of our design over the baseline softmax loss and several state-of-the-art alternatives.