13 January 2020 in IEEE Transactions on Image Processing
Zero-shot learning (ZSL) is a challenging task due to the lack of unseen class data during training. Existing works attempt to establish a mapping between the visual and class spaces through a common intermediate semantic space. The main limitation of existing methods is the strong bias towards seen class, known as the domain shift problem, which leads to unsatisfactory performance in both conventional and generalized ZSL tasks. To tackle this challenge, we propose to convert ZSL to the conventional supervised learning by generating features for unseen classes. To this end, a joint generative model that couples variational autoencoder (VAE) and generative adversarial network (GAN), called Zero-VAE-GAN, is proposed to generate high-quality unseen features. To enhance the class-level discriminability, an adversarial categorization network is incorporated into the joint framework. Besides, we propose two self-training strategies to augment unlabeled unseen features for the transductive extension of our model, addressing the domain shift problem to a large extent. Experimental results on five standard benchmarks and a large-scale dataset demonstrate the superiority of our generative model over the state-of-the-art methods for conventional, especially generalized ZSL tasks. Moreover, the further improvement of the transductive setting demonstrates the effectiveness of the proposed self-training strategies.
1 August 2022 in Computer Standards & Interfaces
25 January 2022 in IEEE Transactions on Multimedia
Zero-shot learning (ZSL) aims to recognize unknown categories that are unavailable during training. Recently, generative models have shown the potential to address this challenging problem by synthesizing unseen features conditioned on semantic embeddings such as attributes. However, unidirectional generative models cannot guarantee the effective coupling between visual and semantic spaces. To this end, we propose a visual-semantic aligned bidirectional network with cycle consistency to alleviate the gap between these two spaces, generating unseen features of high quality. More importantly, we incorporate two carefully designed strategies into our bidirectional framework to improve the overall ZSL performance. Specifically, we enhance the intra-domain class divergence in both visual and semantic spaces, and in the meantime, mitigate the inter-domain shift to preserve seen-unseen domain discrimination. Experimental results on four standard benchmarks show the superiority of our framework over existing state-of-the-art methods under both conventional and generalized ZSL settings.
1 January 2025 in IEEE Journal of Biomedical and Health Informatics
Respiratory diseases are a major cause of death globally, placing a significant burden on healthcare services. Early-stage clinical decision-making is crucial for enabling personalized, prioritized treatment and more efficient allocation of healthcare resources. Clinicians can intervene proactively and develop appropriate treatment plans for patients when provided with vital information such as mortality prediction, deterioration detection, and length-of-stay prediction. To precisely predict such vitals, it is essential to leverage sequential information that is inherent in clinical variables. In this paper, we employ a unified framework for patient outcome forecasting in pneumonia patients. The proposed model utilizes clinical time-series data of varying lengths, along with static admission information, to effectively capture the sequential information of clinical variables. Additionally, we model the imbalanced distribution of mortality prediction and deterioration detection through weight constraints, and we account for the right-skewed distribution of length-of-stay data to enhance the robustness of the model. Furthermore, we develop a data splitting strategy to track dynamic changes in model performance at different timestamps, helping to bridge the gap between testing conditions and real-world scenarios. We conduct experiments on CAP-AI dataset that was obtained and collected from the University Hospitals of Leicester with the involvement of clinicians. It is based on real-world clinical data from patients admitted with pneumonia-related diagnoses. Extensive experimental results demonstrate the effectiveness and robustness of our approach whilst predicting patient outcomes in a clinical setting.