71 results about "Approximation algorithm" patented technology

The invention discloses a multi-modal fusion lithium iron phosphate battery thermal runaway early warning method and system, and the method comprises the steps: collecting multi-source heterogeneous data, such as temperature, voltage, gas concentration and shell strain pressure, in real time through deploying a heterogeneous sensor network, and carrying out the noise reduction and time sequence feature extraction through employing a sub-linear time low-rank approximation algorithm of a Hankel matrix; constructing a cross-modal feature association network by applying a secondary time algorithm of a maximum weight sparse subgraph problem, inputting a fused feature vector into a Bayesian network health degree evaluation model for probabilistic reasoning calculation to obtain a battery health degree score and a thermal runaway risk level, generating a graded early warning signal through multi-level early warning threshold comparison, and performing early warning on the battery health degree score and the thermal runaway risk level. And corresponding prevention and control suggestions are matched. The method solves the technical problems that single physical quantity monitoring is difficult to comprehensively reflect the complex change in the battery and the response delay of a centralized processing architecture causes the early warning lag, and achieves the timely capture and accurate early warning of the early weak characteristics of thermal runaway.

The invention relates to the field of power system planning, in particular to a micro-grid boundary quantitative evaluation method and system for multi-dimensional analysis and dynamic verification. The method comprises the steps that power distribution network and micro-grid scheme data are acquired, and scene recognition and decoupling modeling are carried out after standardization processing; the method comprises the following steps: extracting end supply-preserving scene data, analyzing cost elements, and generating a critical cost threshold table through normalization processing and a threshold approximation algorithm; performing multi-dimensional parameter correlation analysis and integrated learning training based on the table, and constructing a multi-dimensional boundary index model; performing benefit matching calculation and green value accounting according to the result to generate an economic benefit decomposition structure; real-time streaming data processing and stability verification are combined to generate an economical efficiency boundary index set; and finally, generating a standardized evaluation file through matrix mapping and weight dynamic adjustment. According to the method, dynamic quantitative evaluation of the economy boundary of the micro-grid is realized, the capacity substitution benefit and the green value are effectively integrated, and an accurate basis is provided for planning decision.

A method for the joint active and passive beamforming and received signal optimization in an ISAC system assisted by dual IRSs is provided. The method jointly optimizes active beamforming at Base Station (BS), reception of sensing signals at the Base Station (BS), and passive beamforming at IRSs, so as to maximize communication sum-rate of users while ensuring that SNR of sensing signals meets a minimum requirement. To address the complex non-convex optimization problem, the method first applies fractional programming to decouple problem, then adopts successive convex approximation algorithm and alternating direction method of multipliers to transform intractable non-convex problem into multiple tractable subproblems, and finally employs an alternating optimization method to efficiently acquire the high-quality suboptimal solutions. The simulation results demonstrate that disclosed scheme exhibits satisfactory convergence and effectiveness, and can significantly improve the performance of IRS-assisted ISAC systems.

The invention belongs to the technical field of wireless optical communication, and particularly relates to a mobile edge computing task unloading and resource allocation method in wireless optical communication assisted by an unmanned aerial vehicle. According to the method, the data model, the queue model and the communication model are constructed, a task unloading strategy is optimized, and efficient and low-power-consumption edge computing resource allocation is achieved; an optimization technology is adopted to decompose a long-term optimization problem into sub-problems of each time slot, and the sub-problems are solved through a continuous convex approximation algorithm and a BFGS algorithm; finally, joint optimization of task unloading, computing resources, transmitting power, task return transmitting power, time slot scheduling and UAV trajectory is realized, and an optimization framework with more fine granularity in multi-dimensional allocation is formed. The characteristics of flexibility of the unmanned aerial vehicle and high bandwidth of wireless optical communication are utilized, the long-term throughput of the system is improved, the long-term power consumption of Internet of Things equipment is reduced, the high real-time performance of task processing is ensured, and the method is suitable for outdoor mobile Internet of Things application scenes with strict requirements for low power consumption and high real-time performance.

The invention discloses a digital twin high-voltage motor-based strong current operation state monitoring and maintenance system, which comprises the following steps: collecting current, temperature rise and vibration signals in motor operation, carrying out high-precision fitting processing on the signals by using a Chebyshev polynomial approximation algorithm, carrying out trend offset detection on a fitting result based on a CUSUM algorithm, and carrying out high-precision fitting processing on the fitting result based on a CUSUM algorithm; therefore, tiny abnormal changes are identified. And meanwhile, a prediction signal is generated in combination with a digital twinborn model, residual calculation is performed on the prediction signal and the fitting signal, fault features are further analyzed, and early warning information is output. Real-time sensing, trend judgment, fault prediction and intelligent early warning of the running state of the high-voltage motor are achieved, and the system is suitable for motor monitoring and maintenance in an industrial scene.

The invention provides an Internet of Vehicles DNN collaborative reasoning method based on a rate splitting multiple access technology. The method comprises the following steps: S1, establishing an edge computing system model; s2, establishing a communication model; s3, based on a rate splitting multiple access strategy, a road side unit RSU divides the information sent to each group of vehicles into a public flow and a private flow, and distributes a precoding matrix for each group; s4, in combination with a given DNN partition strategy and an RSU computing resource allocation method, calculating the communication rate received by each vehicle by using a continuous convex approximation algorithm; s5, on the basis of the communication rate and RSU computing resources, constructing a directed acyclic graph model of a DNN reasoning task, expanding the model into a network flow graph, and performing task partitioning on the model by adopting a genetic algorithm to minimize reasoning time delay to obtain a corresponding edge-vehicle partition node set; s6, converting an optimization problem into a KKT condition, and solving an optimal allocation scheme of RSU computing resources; and S7, repeating the steps from S4 to S6, and carrying out iterative updating to obtain an optimal solution of the system.

The invention discloses a construction method and evaluation method based on a vehicle network interaction multi-dimensional evaluation model, and the method comprises the steps: automatically recognizing all influence factors in a region through a machine learning algorithm based on intelligent power grid big data and geographic information data, and determining region-level evaluation parameters; fusing the evaluation parameters by a multi-dimensional data fusion method based on deep learning to obtain a comprehensive evaluation index; then, the number of input variables is reduced through feature selection and dimension reduction technologies, meanwhile, an approximation algorithm is adopted to replace complex mathematical operation, and a lightweight assessment model is constructed; and constructing a vehicle network interaction multi-dimensional evaluation model based on the lightweight evaluation model in combination with resource value information and multi-scene requirements. According to the method, the functions of differentiated scheduling, peak clipping and valley filling, frequency adjustment and emergency response are realized, the problems of resource mismatching, scheduling lag and poor regional adaptability caused by traditional static assessment are solved, the power grid stability, the scheduling efficiency and the user participation enthusiasm are improved, and the application of the V2G technology is promoted.

The invention discloses a dynamic beam hopping and resource allocation method for NGSO satellite security communication, and belongs to the field of sixth-generation mobile communication security communication and wireless resource allocation. According to the method, significant non-uniformity of distribution of ground flow requirements in a time domain and a space domain is considered, and a satellite-ground security communication network dynamic beam hopping and resource allocation problem model is constructed by considering ground user flow requirements and composition elements and channel characteristics of an NGSO multi-beam satellite network. Through joint optimization of satellite hopping beam scheduling, power resource allocation and auxiliary interference unmanned aerial vehicle deployment strategies, service requirements of different users and ground eavesdropping environments are dynamically adapted, and the safety throughput and queue delay fairness of the system are improved. According to the method, mixed integer linear programming modeling is adopted, and a low-complexity approximation algorithm is combined, so that the method not only has optimality guarantee, but also can be deployed and operated in an actual satellite communication system, and thus unification of secure communication and efficient resource scheduling is realized.

This invention discloses a multi-clinical-staging disease auxiliary classification method and system based on federated learning, belonging to the fields of big data and medical technology. To improve the accuracy of the classification model and the security of privacy data, this invention collects case data to construct a case dataset. This dataset is input into a client for analysis and processing to obtain an optimized case dataset. The client trains an XGGridBoost model on the optimized dataset. The trained model parameters are then encrypted and compressed, and sent to a central server for decompression. The decompressed model parameters are then securely aggregated, and an approximate algorithm is used to determine the optimal split point. The central server sends the optimal split point to multiple clients. Upon receiving the split point, each client determines whether its local decision tree has reached its maximum depth and performs iterative training or outputs the trained model. This invention solves the problem of data silos and improves the security of privacy data.

The application discloses an automatic planning method and device for a walking route of a land controllable seismic source, and belongs to the field of physical exploration. The application introduces a TSP idea, and finally obtains a seismic source vehicle driving route which is short in path, less in barrier crossing, less in driving restriction, and in line with the driving operation habit of a driver through steps of constructing a map, setting relevant parameters and weight ratios, and solving the walking route of the land controllable seismic source by using a TNH combination algorithm. The application combines the advantages of a two-approximation algorithm, a Heuristic algorithm and an NSGA_II algorithm to propose the TNH combination algorithm for solving the walking route of the land controllable seismic source, so that the automatic planning of the walking route of the controllable seismic source is realized, the operation efficiency is accelerated, the accuracy of operation results is improved, the calculated route is more in line with the demand of field construction, and the exploration operation is further developed intelligently. The application is applied to the field of physical exploration to plan the walking route of the controllable seismic source.

The application discloses a fluid antenna assisted ISAC system joint beam forming and position optimization method based on a quantum particle swarm optimization algorithm. Specifically, an ISAC multi-user system model equipped with a two-dimensional fluid antenna is established, and a beam forming optimization problem is constructed; a weighted least mean square error algorithm is used to convert a communication and rate target function, so that a complex fractional problem is converted into an integral problem, and solving is simplified; an alternating optimization algorithm is used to decompose the original joint optimization problem into multiple sub-problems; an auxiliary variable and a receiving beam former are alternately solved, an iterative solution of a beam forming vector is solved by using a successive convex approximation algorithm, and a fluid antenna position problem is solved by using a quantum particle swarm optimization algorithm. The method of the application effectively improves the communication and rate under the restrictions of the fluid antenna position, the minimum perceived beam pattern gain and the maximum base station transmission power, and is superior to a traditional ISAC system equipped with a fixed antenna.

The present application relates to the technical field of visual domain analysis of digital elevation model, and particularly relates to a DEM visual domain analysis method and system based on volume shadow, which comprises the following steps: S1, obtaining digital elevation model DEM data and user input viewpoint parameters, and constructing a regular triangulated irregular network model based on the DEM data; S2, calculating the yin and yang properties of each triangular face based on the viewpoint parameters, and determining the shadow edge according to the yin and yang properties of adjacent triangular faces and the horizontal distance to the viewpoint. In the present application, the DEM visual domain analysis problem is converted into a volume shadow problem with the viewpoint equal to the light source, and the interpolation processing of the terrain by the DEM data and the GIS platform is neat and standard, which simplifies the calculation related to the volume shadow, realizes the precision of the accurate algorithm, approaches the operation speed of the approximate algorithm, and can consider the inclined surface generated by the interpolation processing of the terrain.

Embodiments of the present application disclose a data processing device and method based on an NPU processor. After determining a target input value, the device and method introduce an auxiliary value, obtain a first calculation value corresponding to the target input value and a second calculation value corresponding to the auxiliary value based on a Taylor expansion, and finally determine a target trigonometric function value from the first calculation value and the second calculation value according to the size of the target input value. Thus, when running the Taylor expansion approximation algorithm in the NPU processor, the accuracy of the value range of the part of the trigonometric function that is difficult to ensure is ensured by processing the auxiliary value corresponding to the target input value.

This invention relates to the field of detection, and in particular to an online method and system for detecting impurities in calcium carbonate ore. Specifically, it includes: acquiring a continuous multispectral image sequence of calcium carbonate ore on a conveyor belt, and performing radiometric correction and spectral normalization preprocessing; identifying dominant spectral clusters of pure calcium carbonate ore by statistically analyzing the pixel spectra of the preprocessed images, and selecting a subset of high-confidence pixel spectra as the initial background training set; training a background dictionary, applying a set of sparse regularization constraints during training, and using the correlation between spectral bands to constrain the atomic energies of the dictionary to continuous or strongly correlated band groups; for each pixel of the detected image, using a sparse approximation algorithm combined with the background dictionary, calculating a sparse reconstruction coefficient vector and a reconstruction residual vector, and obtaining the reconstruction residual and coefficient activation uncertainty measure, respectively; if the reconstruction residual exceeds a first threshold, or the uncertainty measure exceeds a second threshold, it is determined to be an impurity point, and the location and quantity are summarized to generate an online impurity distribution map.

The application relates to a direction modulation-based covert secure transmission method and a communication base station, and belongs to the technical field of wireless communication. In view of the case that channel state information is non-ideal and distortion noise exists, a robust direction modulation technology is adopted, a continuous convex approximation algorithm is used to optimize and select a transmitting end, a combination with low secrecy capacity is eliminated, the secrecy capacity is improved, the possibility of eavesdropping by a eavesdropping user is reduced to a certain extent, and physical layer secure communication is further ensured.

The application belongs to the technical field of artificial intelligence and digital image processing, and discloses an extensible farmland boundary regularization simplification method, system, device and terminal, determines a farmland category list to be segmented; uses an existing labeled training segmentation model to obtain segmentation masks of all farmlands; performs binaryzation processing on the segmentation masks to obtain binaryzation images; performs digital image processing on the binaryzation images to extract the contours of each connected domain; uses the contour information obtained through image processing to perform polygon approximation and remove redundant points to obtain farmland boundary representation. The processing method for extracting local farmland regular boundary uses the farmland segmentation result obtained through deep learning segmentation, uses a traditional image processing method and is supplemented with a polygon approximation algorithm to perform regularization processing on the result, and simultaneously simplifies the contour points of the farmland, so that the number of farmland boundary points and boundary description accuracy can be better balanced while reducing labor cost.

The application belongs to the field of precise batching, and specifically discloses an intelligent precise batching method, which comprises the following steps: acquiring the number of saggers and the single-sagger load; acquiring the batching quantity of each material in multiple materials through an approximation algorithm; according to the adding sequence of the materials, performing actual feeding of a first added material, and determining the actual batching quantity of the first added material according to the batching quantity of the first added material calculated in the previous step; determining the batching quantity of a second added material according to the actual batching of the first added material; performing actual feeding of the second added material, and determining the actual batching quantity of the second added material according to the batching quantity of the second added material calculated in the previous step; repeating the steps S4-S5, and sequentially performing actual feeding of a third added material, a fourth added material, and so on until actual feeding of a last added material; and according to the batching proportion, feeding each material into a mixing equipment area in batches according to a set number of times.

The application relates to a communication and perception integrated multi-base station cooperative anti-shielding millimeter wave waveform design method, belonging to the field of wireless communication and perception. The application realizes the method as follows: a millimeter wave orthogonal frequency division multiplexing signal model is constructed, a user communication rate model under a random shielding environment is constructed, a millimeter wave orthogonal frequency division multiplexing signal perception performance model is constructed, and a base station cooperation perception model in the random shielding environment is constructed; based on the user communication rate model under the random shielding environment, the base station cooperation perception model in the random shielding environment and performance indexes, according to optimization theory, an anti-shielding orthogonal frequency division multiplexing communication and perception integrated waveform design problem based on multi-base station cooperation is constructed; an anti-shielding random successive convex approximation algorithm is constructed, an optimization problem of the communication and perception integrated millimeter wave waveform design is solved, and thus an anti-shielding orthogonal frequency division multiplexing communication and perception integrated beamforming result based on multi-base station cooperation is obtained, and shielding robust communication and perception are realized according to the beamforming result.

The application discloses a microwave ion source multi-pin automatic impedance matching method and device, and relates to the field of microwave ion source technology.The microwave ion source multi-pin automatic impedance device comprises a microwave transmission line, pins, a reflected power detection unit, a driving mechanism and a controller, one end of the pin is inserted into the microwave transmission line and is driven to adjust the insertion depth by the driving mechanism, each pin moves independently, the reflected power detection unit is used for detecting the reflected power from the direction of the plasma load, and the controller is configured to execute the microwave ion source multi-pin automatic impedance matching method.Through the configuration of a plurality of pins capable of coordinated movement on the microwave transmission line, the reflected power detection unit capable of automatically detecting the reflected power and the controller executing a sequential-cyclic approximation algorithm, the plurality of pins are sequentially subjected to local optimal search, and multiple rounds of iteration are performed until the system global optimization, so that the reflected power is finally suppressed to the lowest level, the automatic matching and dynamic tracking are realized, and the optimal matching state is dynamically maintained during the working process.

The invention discloses an antenna radiation characteristic analysis method based on CAT and EI-MOM-PO algorithms, a medium and equipment, and belongs to the technical field of electromagnetic simulation, a target area is divided into an MoM area and a PO area, discretization is carried out, and an RWG primary function is defined. And calculating initial current coefficients of the MoM region and the PO region at the sampling frequency point, introducing correction voltage to update excitation voltage of the matrix equation of the moment method, further solving to obtain a corrected current coefficient of the MoM region, updating the correction voltage, and performing iterative solution until a stable current coefficient of the MoM region is obtained, thereby obtaining a stable current coefficient of the PO region. Current coefficients of the MoM area and the PO area at any frequency point in the whole bandwidth are obtained through a Chebyshev approximation algorithm in the whole bandwidth, and then the radiation characteristics of the antenna are obtained through solving. According to the method, the total simulation time can be greatly shortened while high calculation precision is kept, and the method has good practicability.

The application discloses a transformer substation switch port flow abnormality diagnosis method and device, high-frequency data of the switch is collected through an SNMP probe, a SCD configuration file is analyzed to establish a mapping relationship between communication nodes and switch ports, and a physical network topology graph is constructed; the number of spanning trees of each service node is calculated as an index of network connectivity contribution degree by using a matrix tree theorem, and the importance of services borne by each port is rated; a sliding time window algorithm is used to extract time sequence characteristics, and a global minimum cut approximation algorithm is combined to establish a dynamic baseline model of each port; a differentiated threshold strategy is used for abnormality detection on ports of different service levels; an original dual covering small cut method is used for abnormal event correlation analysis, and a three-layer correlation model of a port-link-protection interval is established. The problems that a traditional static threshold model cannot adapt to dynamic changes of services are solved, and accurate mapping from network abnormality to service influence is realized.

The application discloses an active quasi-zero stiffness construction method based on nonlinear characteristic digital fitting, and steps include: a quasi-zero stiffness system completes automatic zero point calibration, a force actuator unit drives a load platform to make a low-speed quasi-static scanning within a preset stroke, displacement and static force signals are synchronously collected, and a full-stroke force-displacement data set of a support assembly is constructed; then, a numerical approximation algorithm is used for fitting the data set, a digital twin mechanical model containing a high-order nonlinear term is established, and parameters are stored in a control unit. When the system is running, a high-frequency displacement detection unit collects instantaneous displacement of the load platform in real time, the control unit calls the model to calculate a current positive stiffness restoring force, generates a negative stiffness compensation instruction, drives the actuator unit to apply a compensation force to offset the positive stiffness, and makes the system present a quasi-zero stiffness characteristic within a working stroke. The method realizes quasi-zero stiffness, has advantages of no friction, parameter self-adaptation, large dynamic range and the like, and can be widely used in scenes such as spacecraft ground microgravity simulation and high-precision constant force assembly.

The invention discloses a nonlinear dynamic scheduling communication method for a maritime distress and safety system, and the method comprises the following steps: S1, maintaining the state of a shore-based station of a whole network, and constructing and updating an interference graph based on a physical propagation model; s2, a dynamic scheduler based on a heuristic algorithm calculates nonlinear dynamic priorities for communication requests of various message types; and S3, allocating resources in real time according to the calculated nonlinear dynamic priority, converting a resource allocation problem into solving a maximum weighted independent set, and generating a TDMA time slot allocation scheme of each frame based on an approximate MWIS algorithm. According to the method, the core problems of low spectrum resource utilization efficiency, high service response delay, difficulty in considering high-reliability safe communication and dynamic service load adaptability and the like caused by the adoption of a static time division multiple access scheduling mechanism in an existing maritime communication system can be solved.

The invention discloses a sum rate optimization method of an MA and RIS combined auxiliary NOMA uplink transmission system, which is suitable for an uplink transmission system consisting of a multi-antenna base station, an RIS and two communication users, and each communication user is provided with a single movable antenna. In order to effectively improve the transmission capability of the proposed system, the invention provides a sum rate optimization method of an MA and RIS combined auxiliary NOMA uplink transmission system. Specifically, the sum rate of the NOMA uplink transmission system is maximized by jointly optimizing a base station to receive a beam forming vector, an RIS phase shift matrix and an MA position. For the provided complex non-convex problem, a positive semidefinite relaxation algorithm is utilized to optimize receiving beam forming and an RIS phase shift matrix of a base station, a continuous convex approximation algorithm is utilized to optimize an MA position, and alternate iteration is carried out until a convergence condition is met. A simulation result shows that compared with a traditional fixed position antenna and RIS combined auxiliary NOMA uplink transmission scheme, the scheme provided by the invention remarkably improves the sum rate of an NOMA uplink transmission system.

The invention discloses an AWG waveform compression transmission algorithm for quantum computing, and relates to the technical field of communication, and the method comprises the steps: S1, intercepting a target quantum waveform into a first type of wave band and a second type of wave band, and setting independent compression storage; s2, setting a segmented approximation algorithm to generate a first-order compression wave band; s3, performing similarity judgment on all first-order compression wavebands and screening out the first-order compression wavebands; s4, executing run-length coding compression on the screened first-order compression wave band, generating a second-order compression wave band, and performing unified packaging to complete target quantum waveform compression; wherein the waveform type of the first type of wave band comprises a constant sequence waveform, a linear sequence waveform and a repetitive sequence waveform. According to the method, through a staged processing strategy of segmenting, shunting and merging, the scale of data participating in complex operation is remarkably reduced, execution of high-complexity operation on total data is avoided, and the speed and efficiency of compression and decompression processing can be improved.

The invention belongs to the technical field of power grid optimization operation, and discloses a distributed resource power distribution network comprehensive treatment method, system and device and a medium, and the method comprises the steps: constructing an interaction mode of a power distribution network distributed resource and a virtual power plant based on a power distribution network active and reactive cooperative regulation and control mechanism of a virtual power plant technology; aggregating the multi-target distributed resources, constructing a heterogeneous resource aggregation model, and converting the heterogeneous resource aggregation model into standardized feasible region representation and cost function representation by using an equivalent external characteristic approximation algorithm; based on the interaction mode of the power distribution network distributed resources and the virtual power plant and standardized feasible region representation and cost function representation, a cooperative regulation and control model is constructed, the cooperative regulation and control model is utilized to carry out comprehensive regulation and control on the multi-target distributed resources, and an optimal scheduling instruction is output, so that multi-target comprehensive treatment of the power distribution network is realized. According to the invention, the regulation intention of the distributed resource main body is effectively reflected, and the high-reliability implementation of the regulation instruction is realized.

The application discloses a laser focusing laser brightness adjusting method, comprising the following steps: in the focusing process, the brightness of the laser is adjusted to the acceptable threshold range through a preset adaptive laser mode adjusting mode; when the brightness of the laser needs to be adjusted again, the brightness of the laser is adjusted to the threshold range through a preset approximation algorithm; in the focusing process, the brightness of the laser is adjusted through a preset laser mode adjusting mode, and the laser mode adjusting mode has an adaptive function, so that the brightness of the laser can be adaptively adjusted; when the brightness of the laser needs to be adjusted again, the brightness of the laser is adjusted through a preset approximation algorithm, and the approximation algorithm is also adaptive adjustment, so that the manual intervention is reduced, the laser focusing sensor can quickly respond in the focusing process, and the adjustment precision is guaranteed.

The application discloses a digital-to-analog converter and a current calibration device and method thereof, the current calibration method comprising: comparing the output current of a to-be-calibrated current source of the digital-to-analog converter with the output current of a reference current source and outputting an error signal; generating a calibration control code according to the error signal by using a successive approximation algorithm; and controlling a calibration current source to inject corresponding compensation current into the to-be-calibrated current source according to the calibration control code, wherein at least one redundant bit is included in the calibration bits of the calibration control code, and the value of the redundant bit is a positive integer less than N, wherein N is the number of calibration precision bits. The successive approximation algorithm with the redundant bit is adopted, even if a single misjudgment occurs during calibration, the previous error can be corrected in subsequent low-bit calibration, and the calibration stability and calibration precision are significantly improved. Further, the gradient alignment algorithm is adopted, so that the output current of the calibrated current source presents staggered distribution, and the linearity and the spurious-free dynamic range of the signal can be improved.

The invention provides a data processing method based on fusion attention and quantization operation and an accelerator, which are applied to the technical field of computers, and comprise the following steps: based on a plurality of computing cores of a heterogeneous multi-core architecture in the accelerator, executing attention processing operation on a plurality of input data according to a target model to obtain an output result; the normalization core is used for performing normalization processing on the input data to obtain normalized input data; the linear layer core is used for performing linear calculation on the normalized input data to obtain target input data; the fusion attention core is used for executing index calculation operation on the target input data based on a shift index approximation algorithm of a shift index approximation unit to obtain index calculation results, and processing time periods corresponding to the multiple correlation index calculation results are partially overlapped. The multiple correlation index calculation results are multiple index calculation results output by the same fusion attention core, and the processing time period represents the duration of converting the input data into the index calculation results.

The invention provides a common excitation amplitude sparse reconfigurable array synthesis method, and relates to an information transmission and processing technology. The method comprises the following steps: 1) constructing a mathematical optimization model through a radiation characteristic requirement and a null requirement of a to-be-synthesized wave beam, and solving the mathematical optimization model by using particularity of conjugate symmetric excitation and a continuous convex approximation algorithm; 2) utilizing a feasible point tracking algorithm, introducing a slack variable to help to solve a feasible point, and overcoming a non-feasible solution caused by non-convex constraint; the method has the advantages that a plurality of radiation directional diagrams can be formed by one sparse array, so that a plurality of different application scenes can be met; the number of array elements is reduced through the design of a sparse array, so that the system cost is further reduced; due to the characteristic of excitation weight conjugate symmetry, optimization variables are remarkably reduced, and the advantages of the method on a large-scale array are highlighted; the DRR constraint is added, so that the dynamic range and the quantization precision required by the attenuator are reduced, and the hardware complexity is simplified.