Redundant variant detection method and system of deep learning model

By dimensionally reducing and clustering the hidden state sequences of deep learning models, a finite state machine is built to identify redundant variants, which solves the problem of low detection efficiency of redundant variants in the prior art, and improves the accuracy and robustness of the deep learning model.

CN119740177AActive Publication Date: 2025-04-01SUZHOU CITY UNIV
6 Cites 0 Cited by

Patent Information

Application Number
CN202510245238.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-04
Publication Date
2025-04-01
Estimated Expiration
2045-03-04

AI Technical Summary

Technical Problem

Existing redundant variant detection methods are difficult to efficiently identify redundant variants, resulting in the inability to accurately evaluate the quality of image or text test case sets, affecting the accuracy and robustness of deep learning models.

Method used

By obtaining the hidden state sequences of each test sample during the training process of the original deep learning model and each variant, dimensional reduction and clustering are performed, and a finite state machine is constructed to identify redundant variants.

Benefits of technology

It improves the accurate identification of redundant variants of deep learning models, reduces the dependence on large-scale samples and manual labeling, and improves the accuracy and robustness of image or text deep learning models.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the field of software testing, in particular to a redundant variant detection method and system of a deep learning model. According to the method, the hidden state sequence of each test sample in an original deep learning model and each variant is obtained, dimension reduction is performed on each hidden state vector, all the reduced hidden state vectors are clustered, each cluster is regarded as an abstract state, and the hidden state vectors are clustered. A conversion sequence of each test sample between abstract states of the hidden state sequence of each model is used as an abstract state trace of each test sample in the corresponding model; according to the abstract state trace of each test sample in the original deep learning model and each variant, constructing a finite state machine of each model; and identifying the redundant variants by using the original deep learning model and the equivalence and inclusion relationship between the finite-state machine languages of the variants. According to the method, the effectiveness of redundant variant detection is improved, and the accuracy of evaluating the quality of the image or text test case set is ensured.
Need to check novelty before this filing date? Find Prior Art