International Business Machines Corporation

Beijing Jiaotong University

Yi Jin

Lei Li

Chun Ma

Li Li Ma

Bo Yang

Xiao Li

Hong Min

Serge Bourbonnais

You-Chin Gene Fuh

Ying Zhou

Miao Zheng

Wan Chuan Zhang

Wei Wang

Chao Li

Dawei Zhang

Xing Yi Chen

Li Li Wang

Jiqiang Liu

Guorong Chen

Pengrui Liu

Xiaoting Lyu

Xiangrui Xu

Li Duan

Yidong Li

Qing Yun Hao

A method for distributing an equity reward for federated learning based on an equity theory includes the following steps: applying Adams' equity theory to federated learning, analyzing, by a participant, all factors invested in a federated task comprehensively, then giving an expected reward for this task, calculating, by the task publisher, the reputation of the participant; participating, by the participant, in each round of a training task using a local data to evaluate data contribution, model contribution, and a waiting-time allowance of the participant, then combining contribution results of the three factors to evaluate the contribution of the participant; after a global model converges, dynamically adjusting weights of the three factors according to an objective function of the equity reward, with a goal that an actual reward of the participant is as close as possible to the expected reward, and obtaining and distributing the actual reward of the participant.

Method for distributing equity reward for federated learning based on equity theory

Systems for replication group partitioning include a workload profiling module configured to analyze historical workload data for a plurality of data elements to identify and categorize one or more transaction patterns; and a recommendation module configured to generate a recommended partitioning of the plurality of data elements into one or more replication groups, based on the one or more transaction patterns, that are optimized toward a partitioning goal.

Replication group partitioning

Methods and systems for replication group partitioning include analyzing historical workload data for a plurality of data elements to generate one or more transaction patterns and generating a recommended partitioning of the plurality of data elements into one or more replication groups, based on the one or more transaction patterns, that are optimized toward a partitioning goal.

A method, system, and/or computer program product queries a distributed database system. In one embodiment, the query method comprises: parsing a query request; generating an access plan for the query request, wherein the access plan specifies therein a database table related to the query request, and the database table has a plurality of copies stored in different database devices respectively; selecting, based on status information of each of the plurality of copies of the specified database table, one copy from the plurality of copies as a target database table; and executing a query operation in the target database table according to the access plan.

Query method for a distributed database system and query apparatus

A database query by using a user-defined function (UDF) is provided. A computer-implemented method for performing the database query by using the UDF includes detecting whether a command for invoking the UDF to transmit a parameter value has been received. Based on detecting that the command for invoking the UDF to transmit the parameter value has been received: a first interface of the UDF is started; the parameter value is received through the first interface; and the received parameter value is transmitted to an external data source. The method further includes returning to detecting whether the command for invoking the UDF to transmit the parameter value has been received.

Database query using a user-defined function

Data query in a share-nothing database includes obtaining a query request and generating an optimized access plan with respect to the query request. The query request relates to external data stored in an external data source and contains a definition for expected distribution of the external data. The data query also includes obtaining data distribution information related to the expected distribution based on the optimized access plan, transmitting the data distribution information to the external data source so that the external data source splits and returns the external data in accordance with the data distribution information, and executing query-related processing of the split external data in accordance with the optimized access plan.

Database query in a share-nothing database architecture

Embodiments of the invention relate to improved input/output efficiency for online analysis processing in a relational database. An aspect of the invention includes selecting a table from a relational database. The table is split by column into sub-tables, with at least one of the sub-tables including at least two columns. Each sub-table is written into a corresponding table of a row-based storage database.

Input/output efficiency for online analysis processing in a relational database

Growing community of inventors

Beijing, China

Yi Jin