81results about How to "Reduce training overhead" patented technology

The invention discloses an intelligent wave beam training method and a precoding system in millimeter wave communication, and the intelligent wave beam training method determines a codebook interval where an optimal wave beam is located through sensing the change of an environment by utilizing the space-time correlation of a channel. Specifically, a Markov decision process (MDP) is used for modeling a problem; solution is carried out based on deep reinforcement learning, a graphical state space is innovatively constructed to effectively utilize structural information of a channel beam space, and a differentiated action space is provided to improve learning efficiency and convergence speed. The change rate of the environment can be sensed, and the beam training interval is adaptively adjusted, so that the beam training overhead is reduced, and meanwhile, the throughput of the system is effectively improved.

The invention provides a method for constructing a deep convolution neural network model, belonging to the field of mode recognition and machine learning. The method comprises the steps of (1) initializing a convolution neural network model, (2) carrying out end-to-end global extension learning on a network until the average error of a convolution neural network system reaches a preset expected value, (3) after global extension, using a cross validation sample to evaluate network performance, and carrying out local extension learning on the network if a recognition rate does not reach the expected value, (4) adding a new incremental end-to-end extension learning branch, realizing the incremental expansion learning of a network structure, and finally realizing the model construction of the deep convolution neural network. According to the method, a neural element can be added according to needs according to a condition of participating in training samples, the extension expansion and incremental expansion of the network structure are realized, the relevance between a data sample and the network model is enhanced, and the network structure adaptive incremental learning is realized.

The invention discloses a channel estimation method for a passive intelligent reflection surface based on deep learning. The channel estimation method is realized by an offline channel estimation stage and an online channel prediction stage. In the off-line channel estimation stage, in an uplink, a user side sends a pilot signal, a base station side controls an IRS to sequentially open passive elements to reflect an incident pilot signal, and the base station side receives the pilot signal and estimates corresponding cascade channel information through adoption of a minimum mean square error method. An equal probability uniform sampling method is adopted to select a small amount of sampling cascade channel information from the estimated cascade channel information, and the small amount of sampling cascade channel information and complete cascade channel information are adopted to construct a new data set; and in the online channel prediction stage, the base station side estimates a small amount of sampling cascade channel information online and inputs the sampling cascade channel information to the trained ResNet network to recover complete cascade channel information. According to the invention, the number of passive elements can be flexibly selected and the residual units of the residual neural network can be set so as to meet the service quality characteristics of different systems and users.

The invention belongs to the technical field of the wireless communication, specifically a channel of a multiuser large-scale MIMO system under a FDD mode, and relates to a multiuser channel estimation method under large-scale MIMO system and DP priority. The method comprises the following steps: realizing the self-adaptive clustering of the multiuser channel by using the Dirichlet distribution feature, and deriving a method capable of intelligent reconstructing the channel H under the variational Bayesian principle. Compared with the ordinary channel estimation method, the training expenditure of achieving the ideal estimation performance is greatly reduced, and the error between the reconstructed channel and the real channel is small. Compared with the traditional method, the method has the advantages that the computational burden is simplified, the operation speed and the operation precision are improved, and the accuracy of the channel estimation is improved.

The invention belongs to the technical field of wireless communication, and especially relates to a method which reduces the antenna training cost of iteration wave beam shaping by utilizing the channel space sparseness in a wireless MIMO communication system. The rapid iteration wave beam shaping method based on channel space sparseness comprises the steps of initialization, LANCZOS iteration, cycled iteration control and the like. In the time division duplex MIMO system, Uplink and downlink channels can be interchanged, iterated antenna training is carried out by introducing the LANCZOS algorithm, the channel space sparseness is utilized, problems in estimating reception vectors in the iteration process is modeled into a sparse reconstruction problem, theory related to compressed sensing further reduce cost of iteration training, and antenna training is completed at an extremely high speed.

The invention discloses a beam training method in a millimeter wave system assisted by an intelligent reflection surface. The beam training method comprises a multicast training scheme and a layered multicast training scheme. The multicast training comprises two stages. In the first stage, a base station and a RIS perform beam scanning at the same time; and the base station carries out beam scanning on a user, simultaneously divides energy to transmit a beam to the RIS, and controls the RIS to carry out beam scanning on the user; and the user determines an optimal scanning result according to the received signal strength. In the second stage, link-by-link beam training is carried out on the optimal scanning result, and an optimal communication link from the base station to the user is determined. On the basis of multicast training, beam scanning in the first stage is changed into layered beam training by layered multicast training, so that the overhead is further reduced. According to the two schemes provided by the invention, the beam training overhead can be reduced while certain performance is ensured.

The present invention discloses a physical layer safety beam forming method based on a linear neural network. In a multi-input and mono-output system, the reciprocity of a channel is employed, a single antenna legal receiving terminal sends a training sequence, and multi-antenna sender security signals and an artificial noise beam forming weight are obtained through training of a neural network. The three processes of the channel estimation, the channel state information feedback and the beam forming design are combined into a reverse training process employing the linear neural network, onlyone training sequence is sent with no need for feedback, and therefore, compared to a traditional method, the method provided by the invention ensures the information transmission safety while reducing the complexity.

A system and method are provided for facilitating regulatory certification of an aircraft system. The invention provides automated completion and submission of all the designee forms, including: 8110-1 Test Inspection Authorizations (TIA); 8100-1 Conformity Inspection Records; 8130-9 Statements of Conformity; and 8130-3 Airworthiness Approval Tags.

In an exemplary method, one or more components that require certification activity are entered. Based on the entered one or more components, appropriate parties to supply certification activity based on the entered one or more components are automatically identified, and one of the identified parties to perform the certification activity is assigned. If desired, the performed certification activity is received from the assigned party. The performed certification activity is complied into a report, and an interactive display of the compiled report is generated for review.

The invention discloses a three-dimensional training code book design method for a millimeter wave communication system and a wave beam alignment method. The three-dimensional training code book design method comprises the following steps of: 1, according to ranges of a resolution and a wave beam space (with reference to the specification), establishing a three-dimensional training code book tree structure with a depth of S and a degree of N; 2, equally dividing (with reference to the specification) into N rectangular regions (with reference to the specification), and denoting as a set (with reference to the specification), wherein (with reference to the specification) has a wave beam set (with reference to the specification) for corresponding, and a root node of the tree structure is a combination of (with reference to the specification) and (with reference to the specification); 3, determining a b node Cs,b in an s layer, and sequentially determining each node (with reference to the specification) in second to S layers, wherein Cs,b is a combination of (with reference to the specification) and a wave beam set (with reference to the specification); and 4, solving a wave beam forming vector (with reference to the specification) which the wave beam set corresponding to each node comprises so as to obtain the three-dimensional training code book, wherein q=1,..., N. The three-dimensional training code book generated by the method can be used for implementing high-accuracy wave beam alignment and channel estimation, and can obviously reduce training cost of the system.

The invention belongs to the technical field of wireless communication, and particularly relates to a method for rapidly obtaining a plurality of beamforming singular vectors at the same time by performing iteration through the channel reciprocity of a time division duplex system in a wireless multiple input multiple output (MIMO) communication system. The invention provides a rapid iteration antenna training method based on an LANCZOS algorithm. The convergence speed of an LANCZOS method is higher than that of a power iteration method, and a plurality of pairs of receiving-transmitting end beamforming singular vectors of multi-stream beamforming can be obtained by only one stage of iteration, so that the overhead of an antenna training stage is reduced exponentially, and antenna training is finished rapidly. The method has great practical significance.

The invention discloses a training sequence design method in a multi-base-station cooperative system with users as centers, and belongs to the field of wireless communication technologies. According to the method, firstly, an incidence matrix A between base stations and the users, an incidence matrix G among the base stations and an incidence matrix H among the users are calculated according to a clustering result of the multi-base-station cooperative system; then, uplink training sequences and downlink training sequences of the multi-base-station cooperative system are designed, so that the expenditure of training is low. An integer addition operation and an integer sorting operation are mainly used, so that complexity is low, and the training sequence design method is suitable for the characteristic of real-time and dynamic changes of the clustering result in the multi-base-station cooperative system with the users as the centers, can timely design and finish the downlink training sequences and uplink training sequences of the system after clustering is finished each time and is beneficial to being practically applied.

The embodiments of the invention provide a bus stop position obtaining method and device. The method comprises the following steps that: feature extraction is performed on the state parameters of theridge line segment of a target bus line and vehicle state data in the ridge line segment, so that the feature data of the ridge line segment can be obtained; and the feature data are inputted into a preset machine learning network model, and whether bus stops exist on the ridge line segment of the target bus line is determined according to the output result of the machine learning network model, wherein the machine learning network model is obtained by training ridge line segment samples with known bus station result labels, the ridge line segment is obtained by dividing the target bus line according to a preset distance, and vehicles in the vehicle state data comprise buses and/or private vehicles. The method can be implemented by selecting feature data conforming to the field of perfectional knowledge on the basis of the positioning data of bus driving and relatively few road network data; the expenditure of machine learning algorithm training is effectively reduced; the performanceof the model is reliable; and the engineering implementation of the method is easy and convenient.

The invention discloses a zero-sample image hash retrieval method based on a visual-to-semantic network, and the method comprises the steps: converting an image feature vector into a semantic featurebased on the visual-to-semantic network, forming a target loss function by adopting the distance between the semantic feature and a class vector semantic feature, and achieving the classification lossand hash loss of semantic feature reconstruction. The optimal hash method on the training data set is jointly solved, so that the hash retrieval method still has feasibility for non-appeared categories, has better generalization and robustness compared with the traditional hash algorithm, reduces the training overhead of the image retrieval model at present when the data types become richer and richer, and slows down the update period of the hash model. The method can be applied to the fields of image retrieval, image tracing and the like in which database pictures are continuously expanded.

The invention discloses a compressed sensing channel estimation method based on block comparison reconstruction. The compressed sensing channel estimation method comprises the following steps: acquiring a channel sub-matrix from an original channel matrix in a matrix block coding mode; performing two-dimensional discrete Fourier transform on the channel sub-matrix to obtain a frequency spectrum matrix; performing compressive sensing on the vector form of the frequency spectrum matrix to obtain a mathematical relationship between an observation value vector and the frequency spectrum matrix; reconstructing the observation value vector through a compressed sampling matching pursuit CoSaMP algorithm, and restoring the frequency spectrum of the channel sub-matrix; selecting appropriate elements and positions in the frequency spectrum matrix through adoption of a block comparison method, and calculating channel parameters corresponding to the effective path. According to the invention, the compressed sensing is carried out on the frequency spectrum of the channel sub-matrix and only the channel parameters on a limited number of effective paths need to be solved, so that the calculation scale is reduced to a great extent, and the length of the training sequence is further shortened and the training overhead is reduced because the channel sub-matrix is obtained by adopting a matrix block coding mode.

The invention discloses a malicious software family classification avoidance method based on deep reinforcement learning, which adopts a reinforcement learning algorithm to create an environment and construct an intelligent agent, and modifies malicious software to be tested by continuously interacting with a target malicious software classifier to finally achieve the purpose of classification avoidance. Compared with other malicious software detection/classification confrontation methods, the method provided by the invention is easier to implement and lower in overhead; according to the malicious software family classification avoidance method based on deep reinforcement learning, malicious software can be effectively modified for various malicious software classification models constructed by using different features, meanwhile, the maliciousness function of the malicious software cannot be damaged, and finally, malicious software family classification is avoided; and finally, the robustness of the malicious software detection and classification model based on the machine learning algorithm is improved, and the defense capability of the antivirus engine is improved.

The invention relates to a differential beam spatial modulation transmission and blind detection method assisted by sending precoding, and belongs to the technical field of wireless communication. According to the invention, in combination with the idea of beam switching, information bits are mapped to a space-time sending block in a beam space by means of a precoding technology. The method comprises: firstly, setting the column number of a differential beam spatial modulation sending block matrix to be equal to the dimension number of a sending signal vector; then, according to a set closurerequirement of differential transmission, constructing a space-time sending block total set and a differential beam space modulation sending block mapping set of a beam space; then constructing a precoding matrix based on a singular value decomposition mode, using the precoding matrix for differential beam space modulation transmission of a transmitting end; and finally performing differential beam space modulation signal blind detection by a receiving end. The differential beam spatial modulation method designed by the invention has good anti-noise performance, effectively improves the reliability of differential transmission of a communication system, and has practicability.

The invention relates to the technical field of wireless communications, and particularly relates to a method for acquiring a plurality of beam forming singular vectors fast by performing iteration through channel reciprocity of a time division duplex system in a wireless multiple input multiple output (MIMO) communication system. A fast iterative beam forming method under the time division duplex system is designed according to the rapid convergence feature and the property of acquiring a plurality of feature vectors during iteration at one stage of an LANCZOS algorithm, so that the overhead at an antenna training stage can be lowered greatly, and a plurality of beam forming vectors of multithread beam forming can be obtained rapidly.

The invention discloses a social recommendation method and system based on a graph convolutional network, and belongs to the field of machine learning and social recommendation. The method mainly comprises the steps of data set construction, neighborhood aggregation, entity representation generation, model training and prediction and the like. According to the invention, two types of information of articles and social friends interacting with a user side are considered at the user side, and two types of information of users and collaborative similar articles interacting with an article side are considered at the article side, so that a user-article interaction network, a user social network and an article collaborative similar network are modeled in a unified network; and the embedding is improved by explicitly modeling the high-order connectivity among the user-article, the user social network and the article collaborative similar network, so that the deep potential interest preference of the user on the article can be captured, the semantic-rich user/article representation is finally generated, and the purpose of improving the recommendation accuracy is achieved.