132results about How to "Improve iteration efficiency" patented technology

The invention discloses a method for selecting the optimal path of a power distribution and utilization communication network based on the genetic algorithm. The method is characterized by including the following steps that first, a unidirectional topology network is established; second, the population is initialized through binary coding; third, the number of iterations is set; fourth, selecting operation is performed; fifth, crossover operation is performed; sixth, mutation operation is performed; seventh, judgment is performed; eighth, the optimal path is output. On the basis of meeting the constraint conditions, the optimal path is efficiently found so that the search efficiency can be improved and the search complexity can be reduced.

The embodiments of the invention provide an automatic driving simulation test system and method. The sensor target information, transmitted by a virtual scene module, of a target virtual scene is obtained through a sensor module in the simulation test system, the state parameters of a virtual vehicle in a simulation test process are obtained through a positioning module, a vehicle actuator modulegenerates an automatic driving instruction aiming at the state parameters according to the sensor target information, a dynamics module controls the virtual vehicle to execute a simulation test corresponding to the automatic driving instruction to generate a simulation test result, then a state feedback module generates test feedback information by adopting the simulation test result, and a closed-loop simulation test is formed. According to the simulation test system, functions executed by different modules in the automatic driving process of the vehicle are simulated and tested, the reliability and the operation efficiency of the automatic driving test can be effectively improved, developers can conduct technical research and development under the support of credible data, and the efficiency of algorithm iteration in mass production is improved.

The invention relates to an artificial seismic wave synthetic method compatible with a multi-damping ratio response spectrum. By adoption of the method, a traditional method is first adopted to generate initial artificial seismic waves and then a time domain adjustment algorithm is adopted to be performed on all periodical control points in turn; an optimization algorithm is adopted to perform fractional step adjustment in a given periodical control point according to the damping ratio number of a target response spectrum so as to realize a minimum root-mean-square error between a multi-damping ratio response spectrum value on the periodical control point of the artificial seismic waves after adjustment and a target multi-damping ratio response spectrum value. The time domain adjustment algorithm is to overlay amplitude modulation simple harmonic time paths on the initial artificial seismic waves, and the amplitude modulation simple harmonic time paths adopt intensity envelope curves which are the same with the initial artificial seismic waves for modulation of the simple harmonic time paths. The algorithm is high in fitting precision and high in calculating speed and is suitable for promotion and application.

The invention discloses a distributed power system state estimation method. A distributed power system state estimation model is established in consideration of distributed generation (DG) characteristics so that the real-time accurate operational state of each node in the power grid can be obtained, and the real-time accurate operational state of the DG can also be obtained; secondly, a preserving non-linear method is introduced to the state estimation, and the truncation error is avoided so that the state estimation is provided with a higher computational accuracy and the iteration efficiency of an algorithm is improved; and finally, the Jacobian matrix and the Hessian matrix of a function is calculated and measured in the preserving non-linear method of the state estimation, the work load of manually calculating a large quantity of differential functions and compiling differential codes is large, and the work is too cumbersome and easy to produce errors so that the distributed power system state estimation method utilizes the automatic differentiation technique to replace the traditional manual compiling differential codes to calculate the Jacobian matrix and the Hessian matrix, the workload of the manual compiling codes is reduced, and the development efficiency of procedures is improved.

The invention discloses an aircraft full-automatic aerodynamic optimization method based on reinforcement learning and transfer learning. The method is used for solving the problem that an existing pneumatic optimization method is prone to falling into local optimization or low in convergence speed, manual intervention is excluded in the final high-precision optimization stage through the optimization method, and the optimization efficiency is further improved. According to the technical scheme, firstly, a reinforcement learning environment based on semi-empirical estimation and high-precisionfluid simulation is established; a reinforcement learning neural network is constructed; a reward function is set, the global optimization capability of reinforcement learning is utilized; in the network training process, optimization experience is extracted from a semi-experience estimation method and stored in network parameters; then, another reinforcement learning neural network is constructed, migration learning is used for migrating the extracted optimization experience to the network, then the network is applied to aerodynamic optimization based on high-precision fluid simulation, andfinally, high-precision design parameters with excellent aerodynamic performance are obtained by training the network. Compared with a background technology method, the method has the advantages thatthe convergence speed is increased, the strong global optimization capability is realized, and the engineering value for high-precision pneumatic optimization is very high.

The invention discloses a universal phase field method for simulating different failure modes of a brittle material. The method comprises the following steps of: obtaining material parameters requiredby a numerical simulation process through a uniaxial tensile test and a uniaxial compression test, substituting the measured material parameters into an established finite element model, carrying outenergy decomposition on the elastic strain energy of the test sample, carrying out variation on the energy, further establishing a universal balance equation and a crack evolution equation so as to obtain an overall control equation, and solving by adopting an iterative method or a distributed decoupling algorithm, so that accurate simulation of the crack propagation process, the crack propagation direction and the crack propagation path of the brittle material under complex stress is realized. According to the invention, the application range of a phase field model is greatly expanded; the simulation precision of the crack propagation process is greatly improved, and the problem of inaccurate prediction of the crack development direction of a traditional model is solved; and the solvingefficiency is improved, and the risk of non-convergence in the iterative solving process is reduced.

The invention discloses a parallel intrusion detection method based on an unbalanced data deep belief network, and the method comprises the steps: reading unbalanced data set data, carrying out the undersampling of the unbalanced data through employing an improved NCL algorithm, reducing the proportion of most types of samples, and enabling the distribution of the data set data to be balanced; optimizing parameters of the deep belief network model on a distributed memory computing platform Spark platform by adopting an improved differential evolution algorithm to obtain optimal model parameters; performing feature extraction on data of a data set, performing intrusion detection classification by adopting a weighted kernel extreme learning machine, training a plurality of weighted kernel extreme learning machines with different structures in parallel through multiple threads to serve as base classifiers, and establishing a multi-classifier intrusion detection model based on adaptive weighted voting to perform parallel intrusion detection. The technical problems that an existing intrusion detection method lacks pertinence for an unbalanced data set and training time is too long can be solved, and the speed of optimizing parameters of the deep belief network model is increased.

The invention discloses an electric vehicle distribution path optimization method supporting a charging and discharging strategy, and the method comprises the steps: firstly building a target functionwhich takes the fixed cost, driving cost, punishment cost and charging and discharging cost of a vehicle as optimization, and then according to the coordinates of a customer point and a charging station, performing path planning by adopting a hybrid algorithm of a genetic algorithm and local search to obtain an optimal path scheme; and finally, according to the optimal path scheme of the electricvehicle distribution path obtained in the step 2, in combination with the time-of-use electricity price, guiding the electricity market to make a charging and discharging decision for profit maximization in a power supply mode of battery exchange of the battery swap station. According to the distribution path optimization method, an improved genetic algorithm (GA-LS) combined with local search isdesigned, the influence of charging and discharging of the electric vehicle on a power grid is considered, and the distribution path optimization method has the advantages of being high in robust stability, high in iteration efficiency, high in solving quality and the like.

The invention relates to a group target identification method based on a high-resolution range-profile and a single-pulse measurement angle. The method is used for radar group target identification. For the disadvantages that the traditional group target identification method requires each sub-target to move relative to a radar and the computation burden is comparatively large, the invention provides the group target identification method based on the high-resolution range-profile and the single-pulse measurement angle. Based on the high-resolution range-profile and the single-pulse measurement angle information, the clustering is accomplished by adopting a mean value drifting clustering algorithm on a two-dimensional plane, and HRRPs of various sub-targets are extracted without utilizingmovement information of the target; the method has no constraint on a condition whether each sub-target is in movement, and a requirement of requiring each sub-target to move relative to the radar inthe traditional method is solved. By utilizing scattering point amplitude information, the method improves the traditional mean value drifting clustering algorithm from two aspects: initial value selection and sample weight, the iteration efficiency of the mean value drifting algorithm is improved, and the problem that the traditional method is large in computation burden is solved.

The invention provides a method of choosing site and setting capacity for micro power grid. The method comprises the following steps: connecting the micro power grid into different positions; configuring different capacities for the energy storage devices; based on the micro power grid reliability evaluation method, calculating the reliability index of the micro power grid and the overall maintenance costs for energy storage configuration and operation; using the reliability index of the micro power grid and the overall maintenance costs for energy storage configuration and operation as a target; and using the connection positions of the micro power grid and the capacities of the energy storage devices as decision components to optimize planning and seek the connection position and the capacity of energy storage devices with the highest reliability index and the lowest overall maintenance cost for energy storage configuration and operation wherein in the calculation of the reliability index of the micro power grid, the load cutting strategy of the micro power grid shall be considered which relates to the significance weight and the position weight of the load point. The method of the invention can evaluate the reliability of a micro power grid more accurately so as to conduct optimized planning to the micro power grid more properly.

The invention discloses a task scheduling method based on a greedy adaptive ant colony algorithm, belongs to the field of cluster intelligent algorithms, and is mainly used for optimizing the execution efficiency and optimization capability of the ant colony algorithm. Firstly, the greedy algorithm is introduced to accelerate the initialization speed of the ant colony algorithm, so that the ant colony algorithm performs iterative optimization on the basis of the optimal solution of the greedy algorithm, and the optimal iterative efficiency of solving is improved; in the execution stage, an efficiency factor and a volatilization coefficient capable of being adaptively adjusted are also added to accelerate the optimization speed of the ant colony algorithm. The efficiency factor enables theselected node to be more reasonable, and the adaptive adjustment mechanism of the volatilization coefficient enables the algorithm to fully utilize the information of the front and back scheduling results to adjust the volatilization coefficient in a targeted manner, thereby adjusting the optimization direction. An ant colony relay is introduced into the ant colony algorithm to solve the problem that task scheduling cannot be completed by a single ant path under the constraint condition.

The invention aims at the characteristics of artificial intelligence, creatively provides a whole-process artificial intelligence competition system based on a local model and cloud feedback and a data processing method thereof. Through the technical scheme of the invention, pyTorch, tensorFloW, Keras, Scikit-Learn, caffe, MXNet, Theano, Torch and other frameworks for deep learning are operated and supported in local equipment (PC and the like) through a set of script commands and codes, thereby realizing automatic detection and installation of development environment dependence. Debugging ofsmall sample data is performed by using a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit) of local equipment (PC and the like), codes are submitted through script commands to carryout cloud GPU training, cloud training logs are displayed in real time, and training results are automatically fed back by multiple platforms; a deep learning development environment is logged in andautomatically installed at a local terminal, dependence on the development environment is automatically detected and installed, the algorithm development process is omitted, and a network structure method is rapidly iterated. Codes are submitted to a cloud GPU for automatic training and evaluation through a convenient script command, a training log is displayed in real time, and a training resultis fed back in time.

The invention discloses a software and hardware partitioning method on the basis of improved brainstorming algorithms. The software and hardware partitioning method includes initializing parameters; initializing cluster centers; starting iteration updating and ranking individuals from small to large according to fitness values; sequentially starting to compute the distance from each individual toeach cluster center from the first ranked individuals; updating optimal individuals in each cluster; randomly selecting an individual from the clusters and generating a new individual; shifting the randomly selected individual towards global optimal individuals by random lengths randomly generating a new individual meeting hardware area constraint conditions and replacing the randomly selected individual with the new individual; completing an iteration updating process; outputting the optimal individuals to be used as optimal software and hardware partitioning schemes. First-rank individuals sorted according to the fitness values are the global optimal individuals. The software and hardware partitioning method has the advantages that cluster modes and individual updating modes are improved, accordingly, the efficiency of each iteration process can be effectively enhanced, premature convergence can be prevented, the global optimization ability can be omitted, the solution quality can beeffectively enhanced, and the convergence speeds can be effectively increased.

The invention relates to the technical field of big data, and discloses an online inquiry medicine opening system, which is used for improving the iteration efficiency of the online inquiry medicine opening system. The online inquiry medicine opening system comprises a hospital guide system, an inquiry system, an ordering system, a direct accounting system, a payment system and a customization system. The hospital guide system is used for recommending a corresponding doctor online inquiry page; the inquiry system is used for carrying out online inquiry in the displayed doctor online inquiry page to obtain electronic prescription information; the ordering system is used for generating medicine order information according to the electronic prescription information; the direct accounting system is used for deducting prescription amount information according to the drug order information and the corresponding drug cost direct settlement rule information to obtain user self-payment amount information; the payment system is used for paying the self-payment amount information of the user; and the customization system is used for performing function customization on each system. In addition, the invention also relates to a blockchain technology, and the electronic prescription information can be stored in the blockchain node.

The invention provides a dynamic method and system for heterogeneous service data processing logic, and the method comprises the steps: applying to a server, enabling a dynamic compiler to be arrangedon the server, receiving a plurality of rule logics sent by a client, and building a complete rule class according to the plurality of rule logics and a preset rule class framework; compiling instruction information sent by the client side is received, and the dynamic compiler loads and reflects the complete rule class into an object instance according to the compiling instruction information; receiving calling instruction information sent by the client, and obtaining data flow data from the object instance according to the calling instruction information; Performing format conversion on thedata stream data to obtain a plurality of fields; enabling a plurality of fields to execute a plurality of rule logics, so that the influence on the current running service can be reduced, the iteration efficiency of the system is greatly improved, and the user experience is improved.

The invention relates to a real-time simulation and test method for a control system of an automatic driving truck. The method comprises the following steps: S100, establishing a vehicle dynamic model; S200, performing a simulation test and a real vehicle test by using a control algorithm; S300, improving an input scene library; S400, improving a vehicle dynamic model library; S500, designing thecontrol algorithm; S600, performing an automatic traversal test by using a Monte Carlo targeting method; and S700, performing the real vehicle test. The stability of the control algorithm in the wholelife cycle of the vehicle can be verified, the stability and the robustness of control algorithm design are improved, the research and development period of the algorithm is greatly shortened, most of control algorithm design and verification are completed in simulation, a large amount of manpower and material resources can be saved and the accident risk of the real vehicle test is reduced.

The invention discloses a multi-system orthogonal signal low-density check code coding/decoding method, mainly aiming to solve the problems of weak anti-jamming capability, large number of decoding iteration times, great operand and poor decoding performance in the prior art. The method comprises the following specific implementation steps: (1) generating a multi-system low-density check matrix; (2) generating a code word; (3) performing orthogonal signal modulation; (4) receiving a message from a variable message processing device; (5) receiving an estimated code word; (6) judging whether a check equation is satisfied or not; (7) judging whether a maximum iteration time is reached or not; (8) ending decoding. In the orthogonal signal modulation, very high bit error rate performance can be achieved through less decoding iteration times, and the anti-jamming capability and frequency spectrum utilization ratio of the entire system are increased. By adopting fast Hadamard transform, the operation efficiency and storage efficiency of a decoding algorithm are increased, and the programming complexity of the decoding algorithm on a programmable device is lowered.

The invention relates to an endoscope image detection method, a device, a storage medium and electronic equipment, and aims to reduce manpower and time for model deployment in an endoscope image detection scene, improve model deployment efficiency and model iteration efficiency and improve endoscope image detection efficiency. The method comprises the following steps: acquiring an endoscope image to be detected; executing a plurality of target tasks on the endoscope image through an endoscope image detection model to obtain a plurality of task detection results corresponding to the endoscope image, the endoscope image detection model being used for executing the plurality of target tasks through the following steps: extracting image features corresponding to the target tasks from the endoscope image through a task feature network, and performing fusion calculation on each extracted image feature and a pre-trained quality control image feature through an interactive feature network to obtain a fusion feature corresponding to each target task, and for each target task, determining a task detection result corresponding to the target task through a target task network according to the fusion feature corresponding to the target task.

The invention discloses an optimal sampling frequency allocation method and an optimal sampling frequency allocation device of a wireless sensor network. According to the method, in the iteration process, a source node issues the data task information e<j>(t+1); a node j receives the data task information e<j>(t+1) form the source node, and in addition, beta (t<+>) is calculated; the node j judges whether the constraint condition is broken through or not according to the e<j>(t+1) and the beta (t<+>); if the constraint condition is broken through, the updating is carried out, the self link information is sent from the node j to the source node, and if the constraint condition is not broken through, the updating is not carried out; and the source node issues the data task information to a forwarding node of a routing path according to the link information. Through the iteration process, the utility function can be continually increased, when the increase trend of the network utility function approaches to zero, the algorithm completes the convergence, and the optimal solution is obtained. Only when the constraint condition is broken through, the source node and a target node can generate updating data packets, the updating is carried out, and the communication cost can be obviously reduced, so the expenditure of the whole iteration process is far smaller than that of the classical optimization algorithm.

The embodiment of the invention relates to the field of unmanned driving and/or robots. The invention discloses a calibration method of a sensor, the method is applied to a moving object, the moving object is provided with a plurality of sensors; the calibration method of the sensor comprises the following steps: for each movement of the moving object, measurement data generated by each sensor isacquired, the acquired starting moment is the moment when the moving object is in a static state before the movement or the starting moment of the movement, and the acquired ending moment is the moment when the moving object stops the movement for a preset duration; a rotation relational expression of each sensor is determined according to the measurement data of each sensor obtained by each motion; and according to the rotation relation of each sensor, a conversion relation between respective coordinate systems of any two sensors is calibrated. By adopting the embodiment, the accuracy of sensor calibration can be improved.