A method, device and storage medium for generating an adversarial sample
A technology against samples and targets, applied in the computer field, can solve the problems of manual modification and low generation efficiency of confrontation samples, and achieve the effect of improving the generation efficiency
Active Publication Date: 2022-08-05
NANJING TRANSWARP INTELLIGENCE CO LTD
View PDF9 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Among them, the construction and selection of the replacement word set is the key step of the adversarial sample generation algorithm, which mainly includes two construction methods of semantic similarity and appearance similarity; in the prior art, one of the construction methods is usually selected to construct the replacement word set; however, When the effect of adversarial sample generation is not good, it is necessary to manually modify the construction method of the replacement word set, resulting in low efficiency of adversarial sample generation
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0026] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for the convenience of description, the drawings only show some but not all structures related to the present invention.
[0027] Before discussing the exemplary embodiments in greater detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although a flowchart depicts various operations (or steps) as a sequential process, many of the operations may be performed in parallel, concurrently, or concurrently. Additionally, the order of operations can be rearranged. The process may be terminated when its operation is complete, but may also have additional steps not included in the figure...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a method, device and storage medium for generating confrontation samples. The method includes: acquiring at least one target adjustment parameter group matching the target word vector of the original corpus, and adjusting the parameter group according to the target word vector and each target , generate at least one replacement word set matching the word set to be replaced; and then according to each replacement word set, generate at least one candidate confrontation sample corresponding to the original corpus; and according to the degree of difference between each candidate confrontation sample and the original corpus, in Obtain intermediate adversarial samples from each candidate adversarial sample; input each intermediate adversarial sample into the target recognition model respectively, and obtain an intermediate adversarial sample that is inconsistent with the model recognition result of the original corpus as the target adversarial sample of the original corpus. According to the technical solution in the embodiment of the present invention, by acquiring the target adjustment parameter group corresponding to the original corpus, a replacement word set corresponding to the original corpus can be automatically constructed, thereby improving the generation efficiency of adversarial samples.
Description
technical field [0001] Embodiments of the present invention relate to the field of computer technologies, and in particular, to a method, device, and storage medium for generating an adversarial sample. Background technique [0002] Adversarial samples can improve the robustness of the machine learning model. By adding the adversarial samples to the training process of the machine learning model, the resistance of the machine learning model to the adversarial samples can be effectively improved. [0003] At present, the existing adversarial sample generation algorithm mainly includes three steps: the construction of the word set to be replaced, the construction of the replacement word set, and the selection and restriction detection. Among them, the construction and selection of the replacement word set is a key step in the adversarial sample generation algorithm, mainly including two construction methods of semantic similarity and appearance similarity; in the prior art, on...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06K9/62
CPCG06F18/2415G06F18/214
Inventor 唐剑飞张燕
Owner NANJING TRANSWARP INTELLIGENCE CO LTD