10results about How to "Slow convergence" patented technology

The application provides a subway station air conditioning water system energy-saving method based on a Modelica model and a genetic algorithm, relates to the technical field of track transportation building energy optimization and intelligent control, and the method comprises the following steps: a simulation model of a subway station air conditioning water system is established in Dymola software based on a Modelica language; the established simulation model is initialized with a population, a genetic algorithm is realized by using MATLAB, a minimum total energy consumption of the air conditioning water system is taken as an objective function, and an optimal control strategy is obtained; a Dymola-MATLAB joint simulation interface is constructed by using Python according to the optimal control strategy; and the optimal control strategy of the subway station air conditioning water system is obtained by iterative optimization through the genetic algorithm according to the joint simulation interface. Through the technical path of "precise modeling-intelligent optimization-closed loop control", the application systematically solves the industry problems of "high energy consumption and difficult control" of the subway station air conditioning water system, and provides an innovative solution for the construction of intelligent track transportation.

The invention relates to a satellite-borne receiving device and a signal receiving method for resisting malicious interference of an LEO satellite uplink. The satellite-borne receiving device comprises a beam forming network which is used for generating an LEO satellite beam pointing to a target area, receiving an uplink signal sent by a user, reducing the dimension of the uplink signal into an N-dimensional link receiving signal and outputting the N-dimensional link receiving signal; and the square phase shifter network is used for performing interference elimination on the received signals of the links, generating processing signals of the corresponding links and outputting the processing signals to the digital processing unit for signal estimation, and the processing signals output by each link are aggregation of the received signals of all the links after phase shifting. According to the invention, the analog front end with a two-stage structure is designed, so that the function decoupling of beam forming and interference elimination is realized, the interference is completely eliminated at the analog front end, and the algorithm complexity is reduced at the same time.

A multi-voltage phase recovery device and method based on a photorefractive effect are characterized in that the photorefractive effect is utilized to replace a non-linear effect under a strong light condition, diffraction images after an optical system under different non-linear effect degrees are collected through a photoelectric detector, a reconstruction algorithm based on non-linear inversion is used, and a multi-voltage phase recovery model is obtained. Reconstructing amplitude and phase information of the sample to be measured; compared with a traditional linear multi-plane phase recovery method, the method has the advantages that the multiple diffraction images can be obtained only by changing the voltage without mechanical displacement, the axial mechanical displacement error and the transverse alignment error of the multiple images can be avoided, and therefore the precision is higher, and the convergence speed is higher.

The application discloses a personalized picture retrieval method in a federated learning scenario. It includes: 1) loading a public picture-text pair dataset to generate a model training dataset; 2) based on the text information in the dataset in step 1), the tokenizer word segmentation tool is used to obtain the mapping relationship between words and Token positions; 3) based on the picture information in the dataset in step 1), the picture tensor is obtained after preprocessing; 4) a unified model is constructed based on the pre-training model CLIP; 5) the unified model is trained using the text and image expressions obtained in steps 2) and 3); 6) image retrieval is performed using the unified model trained in step 5), and pictures consistent with the query text are obtained. The application designs a private retrieval database for the federated learning client, integrates client-specific information into the pictures and text, and solves the problems of retrieval accuracy decline and slow convergence caused by the non-independent and identically distributed characteristics of data in the federated learning scenario.

The application discloses a polar code construction method based on genetic algorithm acceleration convergence, which comprises the following steps: step 1, setting parameters; step 2, initial genetic algorithm population; step 3, selecting two parents from the population by using a roulette algorithm; step 4, crossing the two parent information bits to generate offspring; step 5, mutating the unlocked information bits of the offspring by using the non-locked frozen bits; step 6, adding the mutated offspring to the population, selecting the optimal one according to the fitness, updating the population and recording the population information; step 7, repeating steps 3 to 6 until the number of genetic iterations reaches a number of spans, judging whether the optimal fitness decreases to a limit or not, if yes, reducing the number of locked bits, temporarily releasing the channel, repeating steps 3 to 6 until the number of genetic iterations reaches a minimum number of spans, if the optimal fitness decreases and the temporarily released channel changes, updating the channel locking condition according to the temporarily released channel, if no, directly repeating steps 3 to 6; and step 8, ending when the genetic iteration stopping condition is met. The application can solve the problem of slow convergence speed by dynamically locking the channel.

The application discloses a kind of off-site computing power dynamic collaborative optimization methods based on multi-agent reinforcement learning, including the following steps: constructing resource topology diagram;Based on resource topology diagram, get node-level state feature vector and system-level state feature vector;Multi-agent environment is constructed;Local observation, global information and agent action set are input into improved CTDE model, output policy network parameters and value network parameters and construct training batch;Based on training batch, get converged policy network parameters and converged value network parameters;Get execution result;Obtain dynamically updated policy network parameters and dynamically updated value network parameters, realize the dynamic collaborative optimization of task acceptance, resource allocation, task migration, replica start-stop and bandwidth matching.

The application discloses a medical document image classification method based on multi-modal fusion, relates to the technical field of medical information processing and artificial intelligence, and realizes automatic recognition and classification of nine types of documents; single-label annotation and consistency review of medical document images are performed based on annotation software; standardization preprocessing is performed on the images, an OCR engine is called to extract text to form an image mode and a text mode; a document image Transformer classification model and a Chinese pre-training language model classifier are trained respectively; under the same hierarchical cross-validation division, two types of features are extracted and late fusion training is performed to obtain a fusion classifier; single, batch and directory level inputs are supported in the reasoning stage, and the output category and confidence are output, and difficult samples can be located through cross-modal divergence analysis to guide data iteration; the application has higher robustness and explainability under long-tail category distribution, and is suitable for medical document automatic archiving, quality control auditing and data governance and the like.

The application discloses a kind of forging step beam servo control system and method, solve the problem of poor performance of prior art step beam control system, with better security, faster response speed Advantage, specific scheme is as follows: a kind of forging step beam servo control system, including industrial computer, controller, data acquisition module, control module, industrial computer is connected with controller, controller is separately connected with data acquisition module, motor control module, motor control module includes driver and motor, driver is connected with motor, motor is connected with ball screw mechanism to distinguish step beam or gripper;Industrial computer is installed with the motion control software for parameter setting and motor motion control program writing;Controller receives instruction from industrial computer and real-time information transmitted by data acquisition module and handles, judges the state and position of motor, compares with motor setting track, state, finally transmits real-time monitoring information to industrial computer.

The invention discloses a method and a system for accelerating a train-ballast track-roadbed system MBD-DEM-FDM coupling simulation model based on a self-adaptive PID (Proportion Integration Differentiation) algorithm. The method comprises the following steps: firstly, establishing a data interaction interface of MBD and DEM-FDM modules, and calculating and transmitting an acting force and a counter-acting force; a self-adaptive PID control algorithm is introduced, loading force is dynamically corrected according to interface errors, and fast approximation balance in a single time step is achieved; the convergence state of the system is monitored through a sliding window error judgment mechanism, a local fuzzy self-learning or global beetle antenna search optimization algorithm is intelligently triggered according to the convergence state, and PID control parameters are dynamically adjusted in real time and globally optimized. According to the method, a traditional time-consuming space iteration balance process is converted into dynamic feedback adjustment of a time domain, the problems that an existing coupling method is low in calculation efficiency, poor in numerical stability and difficult to apply to a large-scale model are effectively solved, and an efficient and reliable simulation tool is provided for design, operation and maintenance of a railway system.

The invention discloses a Beidou navigation satellite real-time rapid high-precision positioning method based on a wind cloud satellite, and the method comprises the steps: collecting the real-time atmospheric precipitable water amount and land surface temperature data of the wind cloud satellite, obtaining a water vapor conversion coefficient through the land surface temperature data, and converting the atmospheric precipitable water amount into zenith wet delay through the water vapor conversion coefficient; then collecting real-time electron density profile data of the wind cloud satellite to obtain the total electron content of an occultation point in the zenith direction, and obtaining the oblique total electron content by adopting a Shepard interpolation method and a mapping relation algorithm; and constructing and training a TimeMixer time sequence prediction model, using the trained TimeMixer model to the zenith wet delay at the current moment, and combining the position of the Beidou receiver, so that the Beidou receiver preferentially adopts the zenith wet delay calculated by the wind cloud satellite to carry out PPP positioning calculation. According to the method, the performance of the low-cost Beidou receiver is remarkably improved, and rapid and high-precision positioning is realized.