A method for generating a new sample base on a generative adversarial network and an adaptive proportion

Abstract

The invention discloses a method for generating a new sample based on a generative adversarial network and an adaptive proportion. The method comprises the following steps: S1, adding noise generatedby the generative adversarial network to the input sample directly and close to the distribution of the input sample; S2, constructing an adaptive proportion according to that variance of the sample,fusing the input sample and the noise generated by the generative adversarial network to generate a new sample, and adjusting the proportion of the noise and the input sample accord to the adaptive proportion of the new sample; S3 supplementing the original sample information for the new sample through the pixel addition operation, and generating a final sample which is beneficial to improving theDNN detection rate. The method improves the accuracy of DNN, and the cost is relatively small, and the complexity is lower.

Description

technical field [0001] The invention belongs to the technical field of deep neural networks, and in particular relates to a method for generating new samples based on generative confrontation networks and adaptive ratios. Background technique [0002] Computer vision mainly uses machines to try to establish an artificial intelligence system that can obtain "information" from images or multi-dimensional data. Specifically, it refers to using cameras and computers instead of human eyes to identify, track and measure targets, and Further graphics processing is done to make the computer processing become an image more suitable for human observation or sent to the instrument for detection. The deep neural network (DNN) is the basis of computer vision applications. In recent years, with the continuous improvement of computer hardware capabilities, the development of DNN technology has also advanced by leaps and bounds, especially in the field of computer vision such as image class...

AI Technical Summary

Problems solved by technology

The optimization problem of a simple neural network is a convex optimization problem. "Convex optimization" refers to a special kind of optimization. Specifically, it refers to a type of optimization problem in which the objective function for finding the minimum value is a convex function. It uses some commonly used algorithms such as Gradient Descend (GD), Stochastic Gradient Descend (SGD), etc. can quickly converge to the minimum value; while the optimization problem of DNN is a non-convex optimization problem, there are a large number of saddle points in the network, adding noise disturbance to the model parameters or according to the Hessian matrix The method of finding the escape direction of the saddle point can improve the performance of the model, but changing the model parameters directly affects the performance of the gradient update algorithm, which may lead to a decline in model convergence, and requires strict theoretical proof of the convergence of the changed optimization algorithm

Images