SELF-LLP: Self-supervised learning from label proportions with self-ensemble. (September 2022)

• Record Type: Journal Article
• Authors: Liu, Jiabin; Qi, Zhiquan; Wang, Bo; Tian, YingJie; Shi, Yong
• Is Part Of: Pattern recognition. Volume 129(2022)
• Publication Date: 2022-09
• Subjects: Learning from label proportion -- Self-supervised learning -- Self-ensemble strategy -- Multi-task learning; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2022.108767
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: A self-supervised learning is introduced to LLP, which leverages the advantage of self-supervision in representation learning to facilitate learning with weakly-supervised labels. A self-ensemble strategy is employed to provide pseudo "supervised" information to guide the training process by aggregating the predictions of multiple previous network evaluations. Although the employed self-supervised mechanism is image-specific, we seamlessly fit our framework to tabular datasets by incorporating orthogonal matrix transformation into the rotation-based self-supervised strategy. Thanks to the self-supervised and self-ensemble mechanisms, our algorithm takes much less epochs to reach convergence and obtain better performance, compared with previous LLP solvers. Abstract: In this paper, we tackle the problem called learning from label proportions (LLP), where the training data is arranged into various bags, with only the proportions of different categories in each bag available. Existing efforts mainly focus on training a model with only the limited proportion information in a weakly supervised manner, thus result in apparent performance gap to supervised learning, as well as computational inefficiency. In this work, we propose a multi-task pipeline called SELF-LLP to make full use of the information contained in the data and model themselves. Specifically, to intensively learn representation from the data, we leverage the self-supervised learning as a plug-in auxiliary task to learn better transferable visual representation. The main insight is to benefit from the self-supervised representation learning with deep model, as well as improving classification performance by a large margin. Meanwhile, in order to better leverage the implicit benefits from the model itself, we incorporate the self-ensemble strategy to guide the training process with an auxiliary supervision information, which is constructed by aggregating multiple previous network predictions. Furthermore, a ramp-up mechanism is further employed to stabilize the training process. In the extensive experiments, our method demonstrates compelling advantages in both accuracy and efficiency over several state-of-the-art LLP approaches.