Multi-task model training method and system based on distributed data

Abstract

The invention discloses a multi-task model training method and system based on distributed data. The method comprises the following steps: circularly executing the following steps until a circular stop condition is met: for any task, determining a storage target data node by a training node; independently training the current local model of the target data node, and determining the gradient of thetask based on the independently trained gradient of the local model loss function of the target data node; the training node summarizes the gradients of the tasks and updates a current local model parameter set of the training node by utilizing a gradient summarizing result; the training node issues a current local model parameter set to data nodes in the system so that each data node in the system updates a local model based on the received model parameter set; and after the cycle is stopped, taking the current local model parameter set of the training node as a model initial parameter set for any task to obtain a specific model for the task.

Description

technical field [0001] The embodiments of this specification relate to the field of model training, and in particular to a distributed data-based multi-task model training method and system. Background technique [0002] In a distributed data storage scenario, data is stored in different data nodes. When performing model training on distributed data, the training node obtains data from different data nodes for model training. For example, when it is necessary to train a classification model for distinguishing illegal users, the training node can obtain user data from multiple banks for model training. [0003] When performing model training for distributed data, the training node can also train specific models with different application scenarios according to different tasks on the basis of the same general model. For example, according to the two tasks of distinguishing violating users and distinguishing risky users, use the same general user classification model to obtain...

AI Technical Summary

Problems solved by technology

[0004] In practical applications, on the basis of the same general model, when different specific models are trained according to different tasks, the model training process of the training node for any task is independent of the model training process of other tasks. Therefore, the overall Inefficient training

Images