A method and system for generating sensor data based on generating countermeasure network

Abstract

The invention relates to a sensor data generation method based on generating an antagonism network. The method comprises the following steps: a model construction step of constructing a generation antagonism network model by using real data through a neural network model, wherein the generation antagonism network model comprises a generator and a discriminator; a model training step of training the generator and the discriminator with a countermeasure game mechanism and performing iteration until the data obtained from the generator satisfies an evaluation criterion; a data generation step ofgenerating composite data by the generator through the countermeasure network model.

Description

technical field [0001] The invention relates to the fields of pervasive computing, deep learning and machine learning, and in particular to a method and system for generating sensor data based on generative confrontation networks. Background technique [0002] Generative adversarial networks (GANs) were first proposed by IanGoodFellow in 2014. The framework of this method is mainly composed of a generation network (generator) and a discriminant network (discriminator). Its goal is mainly to assist training through the discriminant network. A Generative Network that Accurately Learns the Distribution Properties of Raw Data. The generation network and the discriminant network continue to improve their performance in the mutual game and confrontation, that is, the generator continues to evolve to generate synthetic data that is closer to real data, and at the same time, the discriminator also continues to evolve to improve its ability to distinguish between true and false data ...

AI Technical Summary

Problems solved by technology

However, each hidden unit of the traditional cyclic neural network only contains simple operations such as Tanh or ReLU, and there are defects such as gradient disappearance / gradient explosion in practical applications, so it is difficult to deal with long-term dependence problems, which leads to its use in the field of sensor data. The development and application of the

Images