CN113381828B - Sparse code multiple access random channel modeling method based on condition generation countermeasure network - Google Patents
Sparse code multiple access random channel modeling method based on condition generation countermeasure network Download PDFInfo
- Publication number
- CN113381828B CN113381828B CN202110639046.8A CN202110639046A CN113381828B CN 113381828 B CN113381828 B CN 113381828B CN 202110639046 A CN202110639046 A CN 202110639046A CN 113381828 B CN113381828 B CN 113381828B
- Authority
- CN
- China
- Prior art keywords
- theta
- countermeasure network
- channel
- generator
- dnn
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/0202—Channel estimation
- H04L25/0224—Channel estimation using sounding signals
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Power Engineering (AREA)
- Electromagnetism (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
A sparse code multiple access random channel modeling method based on a conditional generation countermeasure network relates to the technical field of information and communication, and aims to solve the problem that channel estimation and decoding of a receiver are difficult due to the fact that multi-user information of an existing method is transmitted to the receiver through different dynamic channels. The invention provides a multi-user-oriented conditional generation-based countermeasure network channel modeling scheme aiming at the problem of dynamic channels in the Internet of things. This scheme takes advantage of the good capture and distribution capabilities of CGAN. CGAN is used to model the channel profile to represent the channel effect. The received DNN-SCMA signal corresponding to the pilot symbols is used as part of the condition information. In particular, the method uses a model-free learning method to accurately learn different types of random channel models, and realizes effective acquisition of dynamic channel information in an actual communication scene.
Description
Technical Field
The invention relates to the technical field of information and communication, in particular to a sparse code multiple access random channel modeling technology for a countermeasure network based on condition generation.
Background
The advent of internet of everything (IoE) networks has created a need for higher spectral efficiency and large-scale internet of things (IoT) terminal connectivity. Non-orthogonal multiple access (NOMA) can effectively meet the above requirements. In recent years, many related techniques have been proposed with respect to NOMA to improve performance. Sparse Code Multiple Access (SCMA) is a typical scheme of code domain NOMA, where the input bits of a user are mapped to a multidimensional constellation by designing an optimized codebook. Due to the difference in the number of resources in different communication environments, conventional codebook designs are limited in actual communication scenarios, and it is therefore necessary to manually construct codebooks for all possible actual communication scenarios.
Deep learning techniques have brought about significant performance improvements in many respects. The method of deep learning is applied to many conventional communication system models, such as channel estimation, channel decoding, etc. In addition, communication systems based on deep learning of unknown channels have been developed, where different types of channel effects can be automatically learned through data-driven methods. As a generative model for deep learning, generative confrontation networks (GANs) can be used to learn the channel Probability Distribution Function (PDF).
Inspired by data-driven approaches, some studies designed a physical layer communication system that employs an end-to-end training strategy to jointly optimize the encoder and decoder networks. The framework based on deep learning can be seen as an Automatic Encoder (AE) system, where the transmitter and receiver are represented by a Deep Neural Network (DNN). Furthermore, deep learning based peer-to-peer communication framework approaches have been applied to SCMA.
Although several DNN-based SCMA (DNN-SCMA) schemes have been proposed from the master learning codebook mapping and decoding strategy, various inherent uncertainties of end-to-end training of the communication system make it difficult for SCMA to build a DNN-based end-to-end communication system for actual channels. In addition, SCMA has the main advantages of providing multi-user access and improving spectral efficiency, but it also means that multi-user information is transmitted to the receiver through different dynamic channels, which causes difficulties in channel estimation and decoding at the receiver.
Disclosure of Invention
The invention provides a sparse code multiple access random channel modeling method for a countermeasure network based on condition generation, aiming at solving the problem that the channel estimation and decoding of a receiver are difficult due to the fact that multi-user information of the existing method is transmitted to the receiver through different dynamic channels.
A sparse code multiple access random channel modeling method of a countermeasure network based on condition generation is characterized in that: it comprises the following steps:
step one, constructing a generator G (z; theta) of the confrontational network G ) A generator G (z; theta G ) The method comprises the following steps: theta G Parameters for the generator representing weights and offset vectors for the generator;
step two, constructing a discriminator D (x; theta) of the countermeasure network D ) D (x; theta.theta. D ) The method comprises the following steps: x is a sample, θ D Representing weights and offset vectors representing the discriminators as parameters of the discriminators;
step three, according to the generator G (z; theta) of the countermeasure network constructed in the step one G ) And the discriminator D (x; theta D ) Generating a sparse code multiple access random channel model for a conditionally generated countermeasure network,
in the condition-based generation of sparse code multiple access random channel model for a competing network:
generator G (z; theta) of the countermeasure network G ) For: will sample P from the distribution z Is converted into analog samples P conforming to the distribution g 。
Arbiter D (x; theta) of the countermeasure network D ) The input D (x; theta D ) Is the actual sample P in the actual data distribution r And a generator G (z; theta) of the countermeasure network G ) Generated simulation samples, the discriminators D (x; theta D ) The output of (c) is a probability.
And completing one-time sparse code multiple access random channel modeling of the countermeasure network based on condition generation.
The invention has the following beneficial effects: the invention provides a sparse code multiple access random channel modeling method for generating a countermeasure network based on conditions. The invention provides a multi-user-oriented conditional generation-based countermeasure network channel modeling scheme aiming at the problem of dynamic channels in the Internet of things. This scheme takes advantage of the good capture and distribution capabilities of CGAN. CGAN is used to model the channel profile to represent the channel effect. The received DNN-SCMA signal corresponding to the pilot symbols is used as part of the condition information. Particularly, the method accurately learns different types of random channel models by using a model-free learning method, realizes effective acquisition of dynamic channel information in an actual communication scene, and improves the accessibility of the traditional end-to-end communication system. End-to-end DNN-SCMA was designed in which the multi-user multidimensional constellation and multi-user decoder of SCMA were autonomously learned using AE architecture. The SCMA's end-to-end system model includes a DNN-based encoder, a CGAN channel model, and a DNN-based decoder. End-to-end training is achieved by using Back Propagation (BP). By iteratively training the constituent networks, end-to-end losses can be optimized in a supervised manner. Simulation results show that a DNN-SCMA that generates a channel model for a counterpoise network based on conditions can achieve similar or better results than the DNN-SCMA without knowledge of channel information. Under the channel of the invention, the difficulty of channel estimation and decoding of a receiver is greatly reduced.
Drawings
FIG. 1 is a block diagram of CGAN-based random channel modeling for end-to-end SCMA for deep convolutional networks;
FIG. 2 is a CGAN-based random channel modeling;
FIG. 3 is a graph of BER performance over AWGN channels;
FIG. 4 is a BER performance under Rayleigh channels;
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
The method for modeling the sparse code multiple access random channel of the countermeasure network based on the condition generation is characterized by comprising the following steps: it comprises the following steps:
step one, constructing a generator G (z; theta) of the countermeasure network G ) A generator G (z; theta G ) The method comprises the following steps: theta G Parameters for the generator representing weights and offset vectors for the generator;
step two, constructing a discriminator D (x; theta) of the countermeasure network D ) D (x; theta D ) The method comprises the following steps: x is a sample, θ D A parameter representing a weight and an offset vector representing the discriminator;
step three, according to the generator G (z; theta) of the confrontation network constructed in the step one G ) And the discriminator D (x; theta D ) Generating a sparse code multiple access random channel model for a conditionally generated countermeasure network,
in the condition-based generation of sparse code multiple access random channel model for a competing network:
generator G (z; theta) of the countermeasure network G ) For: will sample P from the distribution z Is converted into analog samples P conforming to the distribution g 。
Arbiter D (x; theta) of the countermeasure network D ) The input D (x; theta D ) Is a real sample P in the real data distribution r And a generator G (z; theta) of the countermeasure network G ) Generated simulation samples, the discriminators D (x; theta D ) The output of (c) is a probability.
And completing one-time sparse code multiple access random channel modeling of the countermeasure network based on condition generation.
The principle is as follows: system model
In the present invention, a Sparse Code Multiple Access (SCMA) based deep convolutional network (DNN) channel profile is modeled with a conditional generation countermeasure network (CGAN). According to an auto-encoder (AE) structure, a DNN-based encoder and a DNN-based decoder are presented, and multi-user information is reconstructed through end-to-end deep learning.
DNN-based encoder
At the transmitting end, radicals are given according to the AE structureIn the encoder of DNN. DNN is used to enable multi-dimensional constellation mapping from J user data streams to K resource blocks, and K<J. DNN-based encoder f for a system e (. Cndot.) and DNN-based decoder f d (. Consists of a basic DNN unit with multiple hidden layers, so a DNN-based encoder f e (. Cndot.) can be viewed as an SCMA codeword generator. Each hidden layer of the basic DNN unit is composed of a weight matrix, a bias vector and an activation function.
The input data to the DNN-based encoder is a data stream r of length M, and
r j representing the data stream of the jth user. Theta e And theta d Are defined as weight and offset vectors of a DNN-based encoder and a DNN-based decoder, respectively, and
a constellation mapping from the jth user's data stream to the kth resource block defined as DNN based.
The data of the k-th resource block encoded by the DNN-based encoder may be represented as
The data streams of J users are encoded by a DNN-based encoder and multiplexed to a data length of N
Channel modeling for conditional generation-based countermeasure networks
In this section, the SCMA DNN-based encoder and DNN-based decoder are coupled by generating a channel model against the network based on conditions. An approximately accurate conditional channel profile is constructed using a countermeasure network to enable the transfer of gradients from the receive side to the transmit side to enable end-to-end training of the DNN-based SCMA.
GAN is a generative model of the countermeasure network, and is mainly composed of generators and discriminators. The parameters of the generator and the discriminator are set to theta respectively G And theta D . In GAN, the generator G (z; θ) G ) Will mainly sample P from the distribution z Is converted into analog samples P conforming to the distribution g . Input D (x; theta) of discriminator D ) Is the actual sample P in the actual data distribution r And generator G (z; theta) G ) The output of the discriminator is a probability of the generated analog sample. If sample x is identified as being distributed P from the actual data r Extracted from (b), then discriminator D (x; theta) D ) The probability obtained is close to 1, otherwise close to 0. During the training process, the generator G (z; theta) G ) Will attempt to generate a similar to actual channel P r Output samples, then a discriminator D (x; theta) D ) Will try to distinguish the actual channels P r The data sum generator G (z; theta) in (1) G ) The data of (1). The optimization objective function of the generator and the arbiter can be expressed as
The data y received by the receiving end can be expressed as:
y＝hx+n (3)
where H represents the channel state vector, which is obtained by sampling the set of channels H, and n is the noise vector of the channel. Obtaining Channel State Information (CSI) is important for decoding, so in a system, received and pilot symbols y p The corresponding DNN-SCMA signal is added as part of the conditional information m to give the input data x and the received pilot data y p In case (3), the output samples conform to the distribution of y. The present invention learns the current distribution of the actual channel output based on CGAN channel modeling.
We propose to model the channel using CGAN and understand its output distribution. And pilot symbol y p The corresponding received signal is added as part of the condition information m and is therefore givenDetermining input data and receiving pilot data y p In the case of (3), the distribution y to which the samples coincide is output.
Generator G (z | m; theta) of CGAN channel G ) And a discriminator D (x | m; theta D ) This information is used as the condition information.
The target function of CGAN is
DNN-based decoder
At the receiving end, the received signal of the kth resource block of the DNN-based receiver may be written as:
a DNN-based decoder will learn to recover the original information from the signal y received from the channel.
The output data of J users to be recovered can be expressed as:
at the receiving end, the binary cross entropy loss function is used for calculating J user original data streams and user recovery data streams
The distance between them. The loss function can be expressed as:
wherein the content of the first and second substances,
the mth data information representing the jth user data stream at the SCMA sender,
and m first data information representing jth user data stream output by the SCMA receiving end.
System training process
This section introduces the training process for the peer-to-peer SCMA system. To obtain the training data set, first, J user data streams are randomly generated and then mapped to K resource blocks by a DNN-based encoder. Received SCMA signal y, received pilot data y p And the originally transmitted data is collected as a training data set.
The primary goal at the receiving end is to train the DNN-based decoder to recover the input signal r at the SCMA transmitting end. The data received at the receiving end comprises pilot symbols y p And receiving the data y, the output being an estimate obtained by a DNN-based decoder
The DNN-based decoder takes the received SCMA data as input and recovers the transmitted data in an end-to-end manner. The loss function is calculated by equation (7) to train the receiver and obtain the gradient of the loss.
In training a DNN-based encoder, the generator of the CGAN channel is used as the channel, end-to-end cross-entropy loss is calculated at the receiver, and then the weights of the DNN-based encoder that propagates the gradient back through the generator to the DNN-based encoder CGAN channel will be updated based on random gradient descent (SGD), while the weights of the CGAN channel and the receiver remain fixed. In each iteration, the generator and the arbiter are iteratively trained. The parameters of the generator and the arbiter are updated according to the loss function of equation (2).
Performance simulation analysis
In this section, the BER performance analysis of the system under different types of channels is given. The weights of the DNN-SCMA are updated using the SGD and the calculated loss gradient is propagated. A rectifying linear unit (ReLU) is used as the activation function. For an end-to-end DNN-SCMA system, gaussian noise is added to the hidden layer to construct an AWGN channel, and an additional equalization layer is used to construct a Rayleigh fading channel. The end-to-end network system consists of a convolutional layer and a dense layer, and the system parameters are shown in table 1.
TABLE 1 System parameters
Table 1 Parameters of system
The BER performance for the AWGN channel is shown in fig. 3. DNN-SCMA based on CGAN channels have similar BER performance as DNN-SCMA and they have significant performance advantages compared to traditional SCMA.
The BER performance under the rice channel is shown in fig. 4. The BER performance of DNN-SCMA based on the CGAN channel model is slightly higher than DNN-SCMA. These demonstrate the effectiveness of the CGAN channel model.
Claims (5)
1. A sparse code multiple access random channel modeling method of a countermeasure network based on condition generation is characterized in that: it comprises the following steps:
step one, constructing a generator G (z; theta) of the countermeasure network G ) A generator G (z; theta.theta. G ) The method comprises the following steps: theta G Parameters for the generator representing weights and offset vectors for the generator;
step two, constructing a discriminator D (x; theta) of the countermeasure network D ) D (x; theta D ) The method comprises the following steps: x is a sample, θ D Representing the weight and the offset vector of the discriminator as parameters of the discriminator;
step three, according to the generator G (z; theta) of the countermeasure network constructed in the step one G ) And the discriminator D (x; theta D ) Generating a sparse code multiple access random channel model for a conditionally generated countermeasure network,
in the condition-based generation of sparse code multiple access random channel model for a competing network:
generator G (z; theta) of the countermeasure network G ) For: will sample P from the distribution z OfThe acoustic vector z is converted into an analog sample P corresponding to the distribution g ；
Arbiter D (x; theta) of the countermeasure network D ) The input to the arbiter is the actual sample P in the actual data distribution r And a generator G (z; theta) of the countermeasure network G ) Generated simulation samples, the discriminators D (x; theta D ) The output of (a) is a probability;
the method uses a model-free learning method to learn different types of random channel models, and realizes effective acquisition of dynamic channel information in an actual communication scene;
the acquisition mode of the channel learning samples in the random channel model is as follows:
a multi-user multi-dimensional constellation and multi-user decoder that autonomously learns SCMA using AE structure; the SCMA end-to-end system model includes a DNN-based encoder, a CGAN channel model, and a DNN-based decoder; end-to-end training is achieved by using a back propagation BP; optimizing end-to-end loss in a supervised manner by iterative training of the constituent networks；Simulation results show that the DNN-SCMA for generating the channel model of the countermeasure network based on the conditions can achieve the effect similar to or better than that of the DNN-SCMA under the condition that channel information is not known; in order to obtain a training data set, firstly, data streams of J users are randomly generated and then mapped to K resource blocks through a DNN-based encoder; received SCMA signal y, received pilot data y p And the originally transmitted data is collected as a training data set;
and completing one-time sparse code multiple access random channel modeling of the countermeasure network based on condition generation.
2. The method of claim 1, wherein a discriminator D (x; θ) for the countermeasure network D ) If sample x is identified as being from the actual data distribution P r Extracted from (b), then discriminator D (x; theta) D ) The probability obtained is close to 1, otherwise close to 0.
3. The method of claim 1 wherein in the step three, there is a generator G (z; θ) for the countermeasure network in generating the sparse code multiple access random channel model for the conditional generation countermeasure network G ) The training process comprises the following specific processes: during the training process, the generator G (z; theta) G ) Will attempt to generate a similar to actual channel P r Output samples, then a discriminator D (x; theta) D ) Will try to distinguish the actual channels P r The data sum generator G (z; theta) in (1) G ) The data of (1).
4. The method of claim 3, wherein the generator G (z; θ) is a generator of the countermeasure network G ) In the training process:
the generator and arbiter optimization objective function can be expressed as:
5. the sparse code multiple access random channel modeling method for conditional generation based countermeasure network of claim 4, wherein after applying the generated sparse code multiple access random channel model for conditional generation based countermeasure network to an actual communication environment,
the data y received by the receiving end is represented as:
y＝hx+n
where H represents the channel state vector, which is obtained by sampling the set of channels H, and n is the noise vector of the channel.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110639046.8A CN113381828B (en) | 2021-06-08 | 2021-06-08 | Sparse code multiple access random channel modeling method based on condition generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110639046.8A CN113381828B (en) | 2021-06-08 | 2021-06-08 | Sparse code multiple access random channel modeling method based on condition generation countermeasure network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113381828A CN113381828A (en) | 2021-09-10 |
| CN113381828B true CN113381828B (en) | 2022-10-28 |
Family
ID=77572854
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110639046.8A Active CN113381828B (en) | 2021-06-08 | 2021-06-08 | Sparse code multiple access random channel modeling method based on condition generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113381828B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113992313B (en) * | 2021-10-25 | 2023-07-25 | 安徽大学 | Balanced network assisted SCMA encoding and decoding method based on deep learning |
| CN114362859B (en) * | 2021-12-28 | 2024-03-29 | 杭州电子科技大学 | Adaptive channel modeling method and system for enhanced condition generation countermeasure network |
| WO2023150943A1 (en) * | 2022-02-09 | 2023-08-17 | Oppo广东移动通信有限公司 | Method for updating wireless channel model, and apparatus, device and storage medium |
| CN115860054B (en) * | 2022-07-21 | 2023-09-26 | 广州工商学院 | Sparse codebook multiple access coding and decoding system based on generation countermeasure network |
| CN117914656B (en) * | 2024-03-13 | 2024-05-10 | 北京航空航天大学 | End-to-end communication system design method based on neural network |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20100050337A (en) * | 2008-11-05 | 2010-05-13 | 삼성전자주식회사 | Method and apparatus for handover between packet switching domain and circuit switching domain |
| US9225412B2 (en) * | 2012-05-02 | 2015-12-29 | Samsung Electronics Co., Ltd. | Communication system with feedback mechanism and method of operation thereof |
| WO2019135019A1 (en) * | 2018-01-02 | 2019-07-11 | Nokia Technologies Oy | Channel modelling in a data transmission system |
| US11557390B2 (en) * | 2018-04-30 | 2023-01-17 | Elekta, Inc. | Radiotherapy treatment plan modeling using generative adversarial networks |
| CN110289927B (en) * | 2019-07-01 | 2021-06-15 | 上海大学 | Channel simulation realization method for generating countermeasure network based on condition |
| WO2021086140A1 (en) * | 2019-10-31 | 2021-05-06 | Samsung Electronics Co., Ltd. | Method for mdas server assisted handover optimization in wireless network |
| CN112787966B (en) * | 2020-12-28 | 2022-06-28 | 杭州电子科技大学 | Method for demodulating antagonistic network signal based on end-to-end cascade generation |
-
2021
- 2021-06-08 CN CN202110639046.8A patent/CN113381828B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113381828A (en) | 2021-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113381828B (en) | Sparse code multiple access random channel modeling method based on condition generation countermeasure network | |
| CN110474716B (en) | Method for establishing SCMA codec model based on noise reduction self-encoder | |
| Ye et al. | Deep learning-based end-to-end wireless communication systems with conditional GANs as unknown channels | |
| CN110942100B (en) | Working method of spatial modulation system based on deep denoising neural network | |
| CN110113288B (en) | Design and demodulation method of OFDM demodulator based on machine learning | |
| CN111712835A (en) | Channel modeling in data transmission system | |
| CN109039534A (en) | A kind of sparse CDMA signals detection method based on deep neural network | |
| CN109728824B (en) | LDPC code iterative decoding method based on deep learning | |
| CN111711455B (en) | Polarization code BP decoding method based on neural network | |
| CN114268388B (en) | Channel estimation method based on improved GAN network in large-scale MIMO | |
| Zhao et al. | Federated meta-learning enhanced acoustic radio cooperative framework for ocean of things | |
| Zhang et al. | Pilot-assisted MIMO-V-OFDM systems: Compressed sensing and deep learning approaches | |
| CN115309869A (en) | One-to-many multi-user semantic communication model and communication method | |
| Van Huynh et al. | Generative AI for physical layer communications: A survey | |
| CN112422208B (en) | Signal detection method based on antagonistic learning under unknown channel model | |
| Wei et al. | Federated Semantic Learning Driven by Information Bottleneck for Task-Oriented Communications | |
| CN116527180A (en) | SCMA method based on CWGAN-GP satellite-ground link channel modeling | |
| CN116405158A (en) | End-to-end communication system based on deep learning under non-Gaussian noise | |
| CN113489545B (en) | Light space pulse position modulation step-by-step classification detection method based on K-means clustering | |
| CN114759997B (en) | MIMO system signal detection method based on data model double driving | |
| CN113992313B (en) | Balanced network assisted SCMA encoding and decoding method based on deep learning | |
| Yuan et al. | Channel estimation and pilot design for uplink sparse code multiple access system based on complex-valued sparse autoencoder | |
| CN113783782B (en) | Opportunity routing candidate set node ordering method for deep reinforcement learning | |
| JP7455240B2 (en) | Methods, systems, and computer programs for optimizing communication channel capacity using Dirichlet processes | |
| CN114745234A (en) | Deep learning MIMO system signal detection method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |