Adversarial Partial Multi-Label Learning
This addresses the challenge of noisy annotations in multi-label learning for researchers and practitioners, representing an incremental improvement with a novel method.
The paper tackles the problem of learning multi-label prediction models from instances with overcomplete noisy annotations by proposing PML-GAN, an adversarial learning model that uses a disambiguation network and a generative adversarial network to enhance feature-label correspondence, achieving state-of-the-art performance in experiments on multiple datasets.
Partial multi-label learning (PML), which tackles the problem of learning multi-label prediction models from instances with overcomplete noisy annotations, has recently started gaining attention from the research community. In this paper, we propose a novel adversarial learning model, PML-GAN, under a generalized encoder-decoder framework for partial multi-label learning. The PML-GAN model uses a disambiguation network to identify noisy labels and uses a multi-label prediction network to map the training instances to the disambiguated label vectors, while deploying a generative adversarial network as an inverse mapping from label vectors to data samples in the input feature space. The learning of the overall model corresponds to a minimax adversarial game, which enhances the correspondence of input features with the output labels in a bi-directional mapping. Extensive experiments are conducted on multiple datasets, while the proposed model demonstrates the state-of-the-art performance for partial multi-label learning.