Computer-implemented method for training a machine learning algorithm

By training the encoder and classifier centrally on a central computing unit and combining agent tasks and class contrast loss functions, the performance degradation of federated learning in time series data is solved, achieving more efficient fault classification and more accurate intra-class dependency detection.

CN122414261APending Publication Date: 2026-07-17ROBERT BOSCH GMBH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-01-16
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing federated learning methods fail to effectively utilize data features in the case of time series data, resulting in performance degradation, especially in industrial applications where it is difficult to efficiently detect intra-class dependencies during fault classification.

Method used

By training the encoder and classifier centrally on the central computing unit and performing local training on the device using the proxy task module, and by optimizing the model representation using the class contrast loss function, more accurate classification can be achieved.

Benefits of technology

It reduces annotation overhead, improves model scalability and adaptability to data drift, and enhances the accuracy and efficiency of fault classification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122414261A_ABST
    Figure CN122414261A_ABST
Patent Text Reader

Abstract

本发明涉及一种用于训练机器学习算法的计算机实现的方法，其中所述机器学习算法用于运行多个设备，所述方法包括以下步骤：‑ 提供并初始化待训练的模型，其中所述模型包括编码器和分类器；‑ 由所述设备提供用于本地训练的本地训练数据，并由中央计算单元提供用于全局训练的全局训练数据；‑ 由所述设备执行所述编码器的本地训练；‑ 由所述中央计算单元聚合经训练的编码器；‑ 由所述中央计算单元对所述分类器进行全局训练；以及‑ 由所述中央计算单元提供经训练的机器学习算法。
Need to check novelty before this filing date? Find Prior Art