Tencent Technology (shenzhen) Company Limited

Tencent Technology Company Limited

Pin Xiao

Yangyu Tao

Jie Xu

Xiaosen Li

Wen Ouyang

Yong Cheng

Ming Huang

Fangcheng Fu

Jiawei Jiang

Bin Cui

Benlong Hu

Lele Yu

This disclosure relates to a method for vertical federated learning. In multiple participation nodes deployed in a multi-way tree topology, an upper-layer participation node corresponds to k lower-layer participation nodes. After the upper-layer participation node and the k lower-layer participation nodes exchange public keys with each other, the upper-layer participation node performs secure two-party joint computation with the lower-layer participation nodes with a first public key and second public keys as encryption parameters to obtain k two-party joint outputs of a federated model. Further, the upper-layer participation node aggregates the k two-party joint outputs to obtain a first joint model output corresponding to the federated model. As such, a multi-way tree topology deployment-based vertical federated learning architecture is provided, improving the equality of each participation node in a vertical federated learning process.

Method and apparatus for vertical federated learning

This application belongs to the technical field of artificial intelligence, and particularly relates to a data processing method, a data processing apparatus, a computer-readable medium, and an electronic device. The method includes acquiring a relationship graph network, the relationship graph network comprising nodes for representing interacting objects and edges for representing interactive relationships between multiple interacting objects; performing core degree mining on the relationship graph network through a device cluster comprising multiple computing devices, and iteratively updating node core degrees of all or some of the nodes in the relationship graph network; pruning the relationship graph network according to the node core degrees to remove some of the nodes and edges in the relationship graph network; and compressing the device cluster to remove some of the computing devices in the device cluster, when a network scale of the relationship graph network satisfies a network compression condition.

Data processing method and apparatus, computerreadable medium, and electronic device

Provided is a graph data processing method, including: acquiring a degree of association of each node in a network graph; splitting the network graph to obtain a dense subgraph according to the degree of association of each node; determining stable nodes in the network graph and coreness of the stable nodes based on the dense subgraph, the coreness of the stable nodes being greater than a preset threshold; obtaining a sparse subgraph in the network graph according to the preset threshold, and remaining nodes in the network graph other than the stable nodes and connecting edges between the remaining nodes; and determining the coreness of each node in the sparse subgraph based on the sparse subgraph and the stable nodes.

Graph data processing method and apparatus, computer device, and storage medium

A method and system for distributed machine learning and model training are disclosed. In particular, a finite asynchronous parallel training scheme is described. The finite asynchronous parallel training takes advantage of the benefits of both asynchronous parallel training and synchronous parallel training. The computation delays in various distributed computation nodes are further considered when training parameter are updated during each round of iterative training. The disclosed method and system facilities increase of model training speed and efficiency.

Distributed machine learning method and system

A method for accessing metadata includes: receiving an authentication information query request from a query engine, and sending the authentication information query request to a global database server; receiving authentication information from the global database server; sending the authentication information to the query engine so that the query engine determines, according to privilege authentication result of the authentication information, whether to access table information. A corresponding server and system are also disclosed. By storing authentication information in global database server and storing table information in one or more segment database servers, distributed storage of metadata is achieved; the metadata accessing server obtains from the global server the authentication information queried by the query engine and obtains from the segment database server the table information queried by the query engine, which reduces metadata access time and increases the speed of sharing and exchange of information in a distributed data warehouse.

Growing community of inventors

Shenzhen, China

Pin Xiao