Federated-PCA on Vertical-Partitioned Data

Abstract

In the cross-silo federated learning setting, one kind of data partition according to features, which is so-called vertical federated learning (i.e. feature-wise federated learning) (Yang et al. 2019), is to apply to multiple datasets that share the same sample ID space but different feature spaces. Simultaneously, the image dataset can also be partitioned according to labels. To improve the model performance of the isolated parties based on feature-wise (i.e. label-wise) results, the most effective method is to federate the model results of the isolated parties together. However, it is a non-trivial task to allow the participating parties to share the model results without violating the data privacy of the parties. In this paper, within the framework of principal component analysis (PCA), we propose a Federated-PCA machine learning approach, in which the PCA method is used to reduce the dimensionality of sample data for all parties and extract the principal component feature information to improve the efficiency of subsequent training work. This process will not reveal the original data information of each party. The federal system can help each side build a common profit strategy. Under this federal mechanism, the identity and status of each party are the same. By comparing the federated results of the isolated parties and the result of the unseparated party through multiple sets of comparative experiments, we find that the experimental results of these two settings are close, and the proposed method can effectively improve the training model performance of most participating parties.

Federated Graph Neural Networks: Overview, Techniques, and Challenges

“…Then, the information from different clients is uploaded to the server to train a global GNN model for a node classification task. FedSGC [94] assumes that there are only two clients without a central server. Graph topology and node features are owned by two clients.…”

“…There are several benchmark datasets developed for GNNs, including citation network datasets, social network datasets, and chemical property datasets. FedGNNs test their algorithms on these datasets [17], [24], [33], [33], [38], [47], [49], [51], [53], [54], [55], [55], [56], [56], [61], [62], [63], [64], [71], [74], [78], [79], [94], [95], [97], [98], [99] with various data partition methods. FedGNNs also explore many GNN applications in a decentralized setting with privacy concerns.…”

“…For the incomplete graph setting, if there are three clients, then each of them gets one of the node features, graph topology, and node labels of graph data [93]. If there are only two clients, then one will get one component of graph data and the other gets the rest [94]. For the complete graph setting, node features are divided evenly to clients with the graph topology retained by all clients [95], [96], [97], [98].…”

A survey on security and privacy threats to federated learning

“…In this setting, [73,225] design PSFT algorithms about linear regression and xgboost based on SMPC. [87,46,38,65,102,233,49,254] consider the design of the PSFT algorithms by using HE. Specifically, [87,233] propose PSFT algorithms on logistic regression.…”

“…[46,38,65] propose the PSFT algorithm about tree-based model. [102,49,254] design PSFT algorithms about neural network. Furthermore, Chamani and Papadopoulos [29] uses TEE to design PSFT algorithm.…”