Adversarial Learning of Label Dependency: A Novel Framework for Multi-class Classification
This work addresses multi-class classification by exploiting label relations, but it appears incremental as it builds on existing GAN methods for a known bottleneck.
The authors tackled the problem of multi-class classification by modeling label dependencies using a GAN-based framework, where the discriminator learns label dependencies and the classifier generates realistic label sets, resulting in improved generalization on MS-COCO and NUS-WIDE datasets.
Recent work has shown that exploiting relations between labels improves the performance of multi-label classification. We propose a novel framework based on generative adversarial networks (GANs) to model label dependency. The discriminator learns to model label dependency by discriminating real and generated label sets. To fool the discriminator, the classifier, or generator, learns to generate label sets with dependencies close to real data. Extensive experiments and comparisons on two large-scale image classification benchmark datasets (MS-COCO and NUS-WIDE) show that the discriminator improves generalization ability for different kinds of models