Learning Discriminative Features Via Weights-biased Softmax Loss

TL;DR

This paper introduces a theoretically grounded approach to determine the optimal size of the fully connected layer in CNNs and proposes a new W-Softmax loss function that enhances discriminative feature learning and decision margin control.

Contribution

It provides a theoretical analysis for the minimum units in the FC layer and introduces the W-Softmax loss to improve feature discrimination and reduce overfitting in CNNs.

Findings

W-Softmax improves classification accuracy on benchmarks.

Theoretical analysis reduces FC layer size and training time.

W-Softmax allows flexible control of decision margins.

Abstract

Loss functions play a key role in training superior deep neural networks. In convolutional neural networks (CNNs), the popular cross entropy loss together with softmax does not explicitly guarantee minimization of intra-class variance or maximization of inter-class variance. In the early studies, there is no theoretical analysis and experiments explicitly indicating how to choose the number of units in fully connected layer. To help CNNs learn features more fast and discriminative, there are two contributions in this paper. First, we determine the minimum number of units in FC layer by rigorous theoretical analysis and extensive experiment, which reduces CNNs' parameter memory and training time. Second, we propose a negative-focused weights-biased softmax (W-Softmax) loss to help CNNs learn more discriminative features. The proposed W-Softmax loss not only theoretically formulates the intraclass compactness and inter-class separability, but also can avoid overfitting by enlarging decision margins. Moreover, the size of decision margins can be flexibly controlled by adjusting a hyperparameter $\alpha$. Extensive experimental results on several benchmark datasets show the superiority of W-Softmax in image classification tasks.