Universal evaluation method for learning ability of intelligent model

Through the intelligent model learning ability evaluation method with a three-level indicator structure, the problem that the existing evaluation methods cannot fully reflect learning ability is solved, and a comprehensive evaluation of the intelligent model is realized, ensuring the reliability and practicality of the evaluation results, and being applicable to the model Development and continuous monitoring.

CN119066383AInactive Publication Date: 2024-12-03THE QUARTERMASTER RES INST OF THE GENERAL LOGISTICS DEPT OF THE CPLA

Patent Information

Application Number
CN202411206608.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure CN119066383A_ABST
    Figure CN119066383A_ABST
Patent Text Reader

Abstract

The invention provides a general evaluation method for the learning ability of an intelligent model, and aims to comprehensively evaluate the effectiveness and reliability of an artificial intelligence model in practical application. The evaluation index system adopts a three-level structure and comprises a first-level index (a learning ability comprehensive evaluation result), a second-level index (multi-classification dimensions of learning ability) and a third-level index (specific index requirements). The second-level indexes cover learning effects, learning efficiency, learning reliability, learning generalization and learning interpretability, and each aspect is further refined into specific third-level indexes. The evaluation process is divided into three stages: evaluation preparation, evaluation implementation and comprehensive analysis. In the preparation stage, proper evaluation indexes are selected, an evaluation data set is prepared, and model training and adjustment are completed. The implementation stage comprises evaluation of learning effect, efficiency, reliability, generalization and interpretability. And finally, in a comprehensive analysis stage, collecting and sorting evaluation results, performing performance analysis, and forming comprehensive evaluation.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Machine learning model evaluation method based on signal identification

    CN117972370A

  • Reconstruction model multi-dimensional performance evaluation method for aerosol chemical components

    CN118538313A

Cited By

  • Automobile NAND FLASH chip test method, device and equipment and storage medium

    CN120949003A