Support vector machine-based trusted decision-making method based on co-neural network

A technology of support vector machine and neural network, which is applied in the field of credible decision-making based on support vector machine (SVM) co-neural network, which can solve problems such as time-consuming and large footprint, and achieve strong robustness and defense , Improve the reliability and credibility of the effect

CN109101999AActive Publication Date: 2018-12-28EAST CHINA NORMAL UNIV
9 Cites 6 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
Publication Date
2018-12-28

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a trusted decision method of a co-neural network based on a support vector machine (SVM). The method comprises the following steps: firstly, a master neural network (M NN) model is designed and trained with a classification accuracy rate as high as possible as a criterion; Secondly, for M-NN is trimmed, compressed or pruned, and Co NN model is designed and trained accordingto the smaller model volume; then, a pixel attack is carried out on the training set, at which time, M-NN and Co-NN will make a different classification result from the original; finally, combined with the correct label of the training set itself, supervised learning and training is performed on the M-NN and Co-NN classification results to ultimately generate a decision-making SVM model. The experimental results show that the method can effectively resist some pixel attacks in the image recognition field and remarkably improve the robustness of the neural network.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention belongs to the field of artificial intelligence and computer vision, and relates to a decision-making method under a multi-neural network, in particular to a trusted decision-making method of a cooperative neural network based on a support vector machine (SVM). Background technique

[0002] Neural Network (NeuralNetwork) is a common regression and classification model in the field of artificial intelligence. Due to its good prediction results, it has been widely used in various scenarios such as computer vision, natural language processing and deep learning.

[0003] Generally speaking, although a neural network with a more complex structure has higher classification accuracy, it also takes up a larger volume and takes more time to train and predict. However, experiments on the CIFAR-10 dataset show that a neural network with a complex structure does not mean that it is more resistant to certain neural network attacks than a neural netwo...

Examples

Embodiment Construction

[0025] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. The process, conditions, test methods, etc. for implementing the present invention, except for the content specifically mentioned below, are common knowledge and common knowledge in this field, and the present invention has no special limitation content.

[0026] refer to figure 1 , the present invention includes: Master-Neural Network main neural network training, Co-NeuralNetwork co-neural network training, decision-making SVM model generation and credible decision-making neural network architecture verification steps. Each step is detailed below.

[0027] like Figure 2-4 As shown, taking the residual neural network (ResidualNeuralNetwork) as an example, the present invention implements operations such as refinement, cropping, and compression on the M-NN, and finally generates the Co-NN. On the one hand, refine residual blocks (Re...