Unlock instant, AI-driven research and patent intelligence for your innovation.

Neural network training method and device based on critical paths

A neural network training, neural network technology, applied in neural learning methods, biological neural network models, neural architectures, etc., can solve problems such as misleading deep neural networks, machine learning security threats, errors, etc.

Pending Publication Date: 2021-01-05
BEIHANG UNIV
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although these noises have no effect on human cognition and object recognition, they can mislead deep neural networks to make wrong decisions, which poses a serious security threat to the practical application of machine learning in the digital and physical worlds

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Neural network training method and device based on critical paths
  • Neural network training method and device based on critical paths
  • Neural network training method and device based on critical paths

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0072] The technical content of the present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0073] Such as figure 1 and 2 As shown, the neural network training method based on the critical path provided by the embodiment of the present invention includes the following steps:

[0074] 101. For each common sample, obtain the sample-level critical path of the neural network;

[0075] 102. According to the sample-level critical path, the model-level critical path of the neural network is obtained by hierarchical aggregation;

[0076] 103. For the key attack path of the neural network, train the neural network.

[0077] In the embodiment of the present invention, firstly, a data set D is established, including N data samples. The data samples are common samples and non-adversarial samples. Such as figure 2 As shown, among them, select three common samples, image x1 of category A, image x2 of category B and...

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

No PUM Login to View More

Abstract

The invention discloses a neural network training method based on a critical path, and simultaneously discloses a corresponding neural network training device based on the critical path. According tothe invention, the model-level critical path of the neural network is found through the sample-level critical path. The propagation and amplification process of noise in the model is revealed throughthe path, and in the training process of the neural network, the robustness of the neural network is effectively improved by limiting the path.

Description

technical field [0001] The invention relates to a neural network training method based on a critical path and a corresponding neural network training device based on a critical path, belonging to the technical field of deep learning. Background technique [0002] In recent years, deep learning has achieved remarkable results in several challenging fields such as computer vision and natural language processing. In practical applications, deep learning is usually applied to large data sets, which inevitably contain a large amount of noise, including adversarial sample noise and natural noise, in these data sets composed of data collected from daily life. Although these noises have no effect on human cognition and object recognition, they can mislead deep neural networks to make wrong decisions, which poses a serious security threat to the practical application of machine learning in the digital and physical worlds. [0003] At the same time, why small noises can cause deep ne...

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
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/04G06N3/08
CPCG06N3/08G06N3/045
Inventor 刘艾杉刘祥龙李恬霖
Owner BEIHANG UNIV