End-to-end communication receiving method based on prototype network

Abstract

The invention discloses an end-to-end communication receiving method based on a prototype network, and belongs to the technical field of end-to-end communication. According to the method, firstly, based on a prototype network algorithm in small sample learning, a prototype calculation mode is improved to a certain extent according to a channel environment full of noise, and the improved algorithm can avoid the influence of the noise to a certain extent; secondly, the classification task is completed by calculating the Euclidean distance between each support sample and all the class prototypes, the Euclidean distance between the test sample and the prototype of the class to which the test sample belongs serves as an optimization target, and parameters of the network are updated and optimized through continuous training. According to the invention, under the condition that only a small amount of sample data is available, the bit error rate performance lower than that of the existing mode can be realized in a relatively short iteration period. According to the invention, under the condition that only a small amount of sample data is available, when enough cycles are iterated, the bit error rate performance lower than that of the existing scheme can be obtained.

Description

technical field [0001] The invention belongs to the technical field of end-to-end communication, in particular to a method for receiving end-to-end communication based on a prototype network. Background technique [0002] End-to-end communication is a deep learning-based implementation of a communication system based on an autoencoder structure. It uses a purely data-driven approach to simulate the entire process of the signal from transmission modulation, through the channel environment to reception, demodulation and identification. , but like all deep learning-based methods, traditional end-to-end communication requires a large amount of available data and a long training iteration period to achieve good performance, while in the wireless communication environment, the available data for training Labeled data resources are scarce, and an existing approach is to use a meta-learning-based MAML algorithm and apply it to the entire algorithmic structure of an end-to-end commun...

AI Technical Summary

Problems solved by technology

[0002] End-to-end communication is a deep learning-based communication system implementation based on the autoencoder structure. It uses a purely data-driven approach to simulate the entire process of signal modulation from transmission, channel environment to reception demodulation and recognition. , but just like all methods based on deep learning, traditional end-to-end communication requires a large amount of available data and a long training iteration period to achieve good performance, while in a wireless communication environment, available training Labeled data resources are scarce, and one of the existing methods is to use the MAML algorithm based on meta-learning and apply it to the entire algorithm structure of the end-to-end communication system. When used, it can achieve better performance than the classic end-to-end communication method in a shorter iteration cycle, but when the iteration cycle is long enough (10000 iteration cycles), the final performance of the method is the same as that of the classic end-to-end communication method. not much difference

Images