Timezone: »

 
Spotlight
Federated Learning with Positive and Unlabeled Data
Xinyang Lin · Hanting Chen · Yixing Xu · Chao Xu · Xiaolin Gui · Yiping Deng · Yunhe Wang

Wed Jul 20 10:45 AM -- 10:50 AM (PDT) @ Ballroom 1 & 2
in Deep Learning/APP:Computer Vision »

We study the problem of learning from positive and unlabeled (PU) data in the federated setting, where each client only labels a little part of their dataset due to the limitation of resources and time. Different from the settings in traditional PU learning where the negative class consists of a single class, the negative samples which cannot be identified by a client in the federated setting may come from multiple classes which are unknown to the client. Therefore, existing PU learning methods can be hardly applied in this situation. To address this problem, we propose a novel framework, namely Federated learning with Positive and Unlabeled data (FedPU), to minimize the expected risk of multiple negative classes by leveraging the labeled data in other clients. We theoretically analyze the generalization bound of the proposed FedPU. Empirical experiments show that the FedPU can achieve much better performance than conventional supervised and semi-supervised federated learning methods.

Keywords: APP: Computer Vision · Deep Learning

Author Information

Xinyang Lin (Xi’an Jiaotong University)
Hanting Chen (Peking University)
Yixing Xu (Huawei Technologies)
Chao Xu (Peking University)
Xiaolin Gui (Xi'an jiaotong university)
Yiping Deng (Huawei)
Yunhe Wang (Noah's Ark Lab, Huawei Technologies.)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors