64 results about "Greedy selection" patented technology

The invention discloses an RBF neural network optimization algorithm based on an improved sparrow search algorithm. RBF initial parameters are optimized through the improved sparrow search algorithm,so that the sea clutter prediction precision is further improved, and the purpose of suppression is achieved. An elite reverse learning strategy is introduced, a current optimal solution is selected as an elite individual, and a reverse solution of the elite individual is generated, so that the global search capability of the algorithm is enhanced. Self-adaptive Gaussian variation is adopted to perform variation on an optimal solution and perform greedy selection, and in addition, a position updating mode for sparrow investigation early warning is also improved. And the population is promotedto evolve towards the optimal solution direction, so that the problem that sparrows are easy to fall into local optimum in the convergence process of low fitness in the sparrow search algorithm is avoided to a certain extent. The ability of the improved sparrow search algorithm to jump out of the local optimum is enhanced, and the convergence speed and precision of the RBF network optimized by theimproved sparrow search algorithm are further improved.

A downlink multi-user multiplexing method integrating with receiving antenna selection and zero-approach bean formation (ZABF) includes setting number of radio frequency component at user end to be less than number of receiving antenna, operating greedy selection algorithm on user channel matrix by user end, selecting line with matrix second-order norm (MSON) from said matrix, projecting the rest lines and selecting projection line with MSON, repeating selection steps till number of radio component selected by user is equal to number of antenna subset, activating channel state information of all users to receive selection of antenna subset and ZABF.

The invention discloses a video abstract method and system used for a self-made video. The video abstract method and system is used for solving the problems of time and labor consumption of generatinga video abstract of the self-made video at present. The method comprises the steps of segmenting an original video by using a dense neighbor clustering method to generate multiple video clips; assessing important levels of the video clips; and sampling the video clips with the important levels in a preset range through a greedy selection policy to generate the video abstract. The video clips aresubjected to the important level assessment through an unsupervised learning method, so that the use of the video abstract generated by continuous detailed and accurate video frames is ensured; and the manually tagged video abstract is not required to assist in analyzing a video internal structure, so that a large amount of labor and time are saved.

The invention provides a CCFD-Massive MIMO system power distribution method based on quantum backtracking search optimization. The CCFD-Massive MIMO system power distribution method based on quantum backtracking search optimization comprises the steps of: establishing a system model; initializing a quantum population and system parameters, so that mapping modes of quantum individuals are obtainedthrough a mapping rule; calculating adaptive values of the quantum individuals, and taking the quantum individual, the adaptive value of which is the maximum in the quantum population, as the globallyoptimal solution; generating new quantum individuals through evolution and crossover strategies; obtaining mapping modes of the new quantum individuals according to the mapping rule, calculating adaptive values, and updating the quantum population and the globally optimal solution through greedy selection; if the iteration time is less than the pre-set maximum iteration time, returning to the fourth step; and otherwise, ending iteration, and outputting the globally optimal solution, so that an optimal power distribution scheme is obtained. According to the CCFD-Massive MIMO system power distribution method based on quantum backtracking search optimization provided by the invention, the spectrum utilization rate is effectively increased; self-interference and mutual interference of base stations and users are sufficiently considered; the secrecy capacity of a system is increased to a great extent; and a new solution is provided for the power distribution problem of a complex system.

A ducted unmanned aerial vehicle anti-sway method based on optimized quadratic form control of artificial bee colony includes eight steps: 1, the mathematical model for pendulum oscillation is built; 2, control structure and control law are designed; 3, the parameters of the artificial bee colony algorithm and the employed bee colony are initialized; 4, the performance indicator function of the linear quadratic form is calculated according to individual parameters; 5, worker bees select bee individuals with better fitness as leading bees according to the fitness value of each employed bee, and each worker bee continues to seek for honey sources near the leading bee solution space and the fitness value is calculated; 6, if the number of searches Bas is greater than the set threshold, the employed bees seek for new honey sources again, namely parameter values are initialized again; 7, by the greedy selection method, the positions of the employed bees are updated with larger fitness values, and searches proceed near solution spaces; and 8, the Step 4 is carried out repeatedly until T > Tmax, and the optimum component weighted value parameter, the optimum feedback gain matrix and the optimum fitness value are output.

The invention discloses a query processing method and system for big data platform materialized views. The steps for selecting the materialized views are as follows: generating an MVPP (Multi-View Processing Plan) structure diagram for a given query collection, obtaining a collection of all nonleaf nodes according to the structure diagram, calculating a value of each nonleaf node in the collection, and obtaining a materialized view collection by employing a revenue maximization oriented materialized view greedy selection algorithm; and the steps for placing the materialized views are as follows: establishing a materialized view associated weight matrix for the materialized view collection obtained at the steps for selecting the materialized views, calculating the value of each element in the matrix, arranging all computational nodes in a descending order according to magnitudes of materialized storage spaces, taking the node with the largest materialized storage space in all computational nodes, and obtaining the materialized views placed at the node. According to the materialized view placing method disclosed by the invention, the network data transmission among the nodes can be reduced, and the processing time is shortened.

The invention discloses an improved genetic algorithm-based travel itinerary planning method. The method comprises the following steps of firstly performing arrangement according to a sequence of visited cities to form codes; secondly initializing a population by adopting a two-way greedy selection policy; calculating a fitness value of each individual in the population; by adopting roulette wheel selection, selecting the individuals with high fitness from the old population to a new population; performing crossover operation according to an adaptive crossover probability Pci, and selecting multiple parents to perform pairing to generate new individuals; performing mutation operation according to an adaptive mutation probability Pmi, and determining mutant individuals; and finally judging whether a predetermined stop condition is met or not, and if yes, stopping heredity and obtaining an optimal solution, otherwise, calculating the fitness value of each individual in the population. According to the method, a travel itinerary route is planned for users by adopting an improved greedy adaptive genetic algorithm based on a travel itinerary planning model; and through the method, the itinerary planning speed is increased and the algorithm is prevented from falling into local optimal solution.

The invention provides a vehicle task allocation method under determined and an undetermined tracks in crowd sensing, which is characterized in that the method is used for ensuring the execution of all tasks, and selecting vehicles to complete the tasks at the lowest cost under a deterministic track; under an uncertain track, firstly, the probability of each vehicle track is determined through a logistic regression method, then, the probability of task execution of each vehicle is calculated through a semi-Markov method, and finally, a proper vehicle is selected to ensure that the probabilityof joint execution of each task is greater than a probability threshold value and the cost is minimum. The emdoiment is that as for an NP-hard problem, a greedy selection-based mode is adopted, and for a given vehicle set V and a given task set S, a deterministic trajectory task allocation process or an uncertain trajectory task allocation process is executed to solve a task set S executed by thevehicle set with the least cost. According to the technical scheme, resources can be saved, and efficient road network traffic state monitoring is supported.

The invention relates to the technical field of neural network optimization, in particular to an RBF neural network optimization method based on an improved Harlisia eagle algorithm, which optimizes RBF initial parameters through the improved Harlisia eagle algorithm and realizes accurate prediction and suppression of sea clutters. According to the invention, the coefficient r3 in the position updating formula in the exploration stage is correspondingly improved, and the balance between the global search capability and the local search capability of the algorithm is fully considered. A part of individuals with poor fitness is selected to carry out non-uniform variation and greedy selection. Besides, the E is segmented to ensure that global search is executed in the early stage of iteration of the algorithm, the global search capability can be kept under a certain probability in the later stage of iteration, and the possibility that a global optimal solution cannot be found due to search stagnation caused by falling into local optimum is reduced. The global search capability of the improved Harris eagle algorithm is enhanced, and the RBF network optimized by the improved Harris eagle algorithm provides more rising space for the sea clutter prediction precision.

The invention provides a near field source direction finding method based on a quantum whale optimization mechanism. A near field source direction finding model is established; whales in a whale groupare initialized, a fitness function is constructed, the fitness of each whale quantum position is calculated, and a global optimal quantum position is determined; each whale selects one from quantumevolution rules including spiral update, a contracted bounding mechanism and random food searching according to the probability to update the quantum position of itself; the updated whale quantum position is mapped to the whale position, the fitness of each whale quantum position is calculated, a greedy selection strategy is used to select the quantum position, and the globally optimal quantum position is updated; and the globally optimal quantum position is output, and maximum likelihood estimated values of the corresponding azimuth and distance are obtained by mapping transformation. The method can be used to carry out high-precision range finding on coherent near field sources, and the quantum whale optimization mechanism is used to realize the high-precision range finding method.

The invention discloses an urban road network optimal restoration sequence scheme based on a greedy algorithm. The method comprises the steps of: a, obtaining a set of roads to be restored in a known road network; b, defining importance discrimination indexes from the aspect of road network restoration; c, calculating importance discrimination indexes of each road in set of roads to be restored; d, utilizing a greedy selection idea to restore the most important road in the set of roads to be restored; e, repeating the steps c and d on residual roads to be restored until the set of roads to be restored is empty because after one road in the road network is restored, the state of the road network is changed, and importance discrimination indexes of the roads change too; and f, outputting a road restoration sequence, wherein the sequence is the optimal restoration sequence, and the obtained restoration sequence scheme is the optimal restoration sequence scheme of the road network. The applicability of the greedy algorithm to the road network optimal restoration sequence problem is theoretically proved, and by adopting the greedy algorithm, a large amount of tedious calculation is avoided, the efficiency is high, and the result has important guiding significance to urban road network optimal restoration in real life.

The invention discloses an interference-suppression and energy-saving cellular network relay station deployment method, which aims at problems that the traditional cellular network has high energy loss, high base station apparatus costs and low energy efficiency. The interference-suppression and energy-saving cellular network relay station deployment method comprises the steps: firstly, according to position distribution of users, determining position coordinates and quantity of candidate relay stations by utilizing a K-means algorithm; drawing an interference figure of the candidate relay stations; constructing a basic model which can maximize energy efficiency under limitation conditions of deployment budget, spectrum efficiency and interference threshold distance; performing relay station optimal deployment selection by utilizing a greedy selection way; finally determining the positions and quantity of the relay stations. The interference-suppression and energy-saving cellular network relay station deployment method can be applied to cellular relay network, and can directly deploy a new relay station on a desired place, thereby reducing the total energy loss of the communication system, and meanwhile, ensuring the system capacity and the user service quality; the interference-suppression and energy-saving cellular network relay station deployment method can be further applied to deploying other sites of the cellular network, such as micro NodeBs and femtocells.

The invention discloses a sparse multiband signal reconstruction method based on conjugate gradient tracking. The method comprises the following steps: firstly, establishing a signal reconstruction model; then constructing a reconstruction framework; and finally using an MMV-CGP algorithm to reconstruct a signal, wherein the algorithm converts an OMPMMV algorithm solution problem into an unconstrained optimization problem, and a greedy selection method similar to the OMPMMV algorithm is adopted and is replaced into pseudo inverse operation by conjugate gradient optimization, and thus the purpose of reducing the computational complexity and the computational cost in a signal reconstruction process is achieved.

The invention discloses a method of increasing electric energy quality based on processing an active optimization constraint condition. The method is characterized by establishing an active optimization mathematics model of an electric power system; generating an initial population and setting an operation parameter; determining a target function including a penalty coefficient; selecting a global optimal solution and updating the penalty coefficient; executing variation and intersection operations in a differential evolution algorithm and generating a new test individual; calculating an individual fitness value and a constraint assessment value; and using a non-greedy selection strategy to select one from the new test individual and an original individual to be served as a new individual of a next generation and updating the global optimal solution. A penalty function and the non-greedy selection strategy are combined so as to ensure that the individual is gathered towards a feasible area to acquire an optimal solution. Through using the method, after electric power system active optimization and the differential evolution algorithm are combined, on an aspect of an electric power system active optimization problem, a good search capability and a good convergence effect are possessed.

The invention discloses a puncturing algorithm of a finite length LDPC (Low Density Parity Check) code. According to the puncturing algorithm, by extracting all nodes with high decoding recovery reliability from an original code, then reconstructing a to-be-deleted sequence having valid check information uniformly distributed, and then inserting the nodes with the high decoding recovery reliability into the to-be-deleted sequence according to a certain rule, the shortage of greedy selection in the traditional grading sorting algorithm is overcome, the loss of low bit rate decoding performance due to excessively uniform distribution is avoided, the overall decoding is improved and the error floor of the overall decoding is reduced.

The present invention proposes a function optimization-oriented social learning method, including selection, crossover, mutation operation, imitation learning operation, observation learning operation, update operation, and influence operation. The evolution process of the operation; the observation learning and imitation learning operations are designed in the learning space; in the observation learning operator, the learning inertia weight and the learning disturbance factor are introduced respectively, taking into account the learning inertia and the learning volatility. , which is helpful to expand the search range of the method and improve the convergence speed of the method; the imitation learning operator introduces a variable search step size to improve the convergence speed of the method. In the belief space, the update function of knowledge and the influence function of culture on microspace are designed. After adopting this method, the solution accuracy can be significantly improved, and the convergence speed is faster.

The invention discloses a method for optimizing spectrum efficiency of a wireless energy-carrying communication system based on hybrid precoding. The method comprises the following steps: S1, obtaining an analog precoding matrix F from a base station BS in the wireless energy-carrying communication system through a step-by-step greedy selection algorithm; S2, enabling the base station BS to maximize the spectrum efficiency of the simultaneous wireless information and power transfer system through a soft interference elimination algorithm under the condition that constraint conditions are met, and obtaining a digital precoding matrix B; S3, transmitting information by the base station BS through the analog pre-coding matrix F and the digital pre-coding matrix B so as to maximize the spectrum efficiency of the simultaneous wireless information and energy transfer system. In the signal transmission process, the service range of the SWIPT can be expanded, certain energy is provided for an energy receiver, and a higher data rate can be provided for a user by using the millimeter wave Massive MIMO technology; and a hybrid precoding structure is adopted, so that the hardware complexity and energy consumption of the system are reduced.

The invention discloses a method for monitoring deployment of a wireless sensor network in a target irregular distribution environment. The method comprises the following steps: I, determining initialdesired sensing ranges of sensor nodes at different positions according to the distance away from a base station; II, determining local density and a local density level of monitoring targets according to a monitoring target distribution situation; III, determining actual desired sensing ranges of the sensor nodes according to the local density of the monitoring targets; and IV, deploying the sensor nodes based on greedy selection and a rebounding way to meet the basic coverage requirement and the load balancing demand. According to the method, the sensor nodes are distributed at higher density in a region closer to the base station and the sensor nodes are distributed at higher density in a region where the monitoring targets are distributed densely to further realize network load balance after the basic coverage requirement of node deployment is met, thereby greatly prolonging the network life.

The embodiment of the invention provides an adversarial sample generation method and device and computer equipment. The method includes: determining associated elements of a target node in the graph data sample; modifying the graph data sample according to the associated elements, inputting the modified graph data sample into a target model, determining an interference parameter representing the interference size of each associated element on the target model according to a result output by the target model, and then selecting the target element with relatively large interference to modify thegraph data to obtain an adversarial sample. The disturbance result of the associated element is accurately quantified through forward calculation of the model, the problems of continuous relaxation in calculation of the disturbance result based on gradient information and inaccuracy of the calculated disturbance result are avoided, and the least disturbance is added to the graph-junction data through a greedy selection strategy to generate an adversarial sample.

The invention provides a gene selection method which is used for gene characteristic selection. The method comprises the steps of obtaining a training set and a test set through a gene data microarraydata set, and determining an initialized population; carrying out binary coding on each individual of the current population by adopting a conversion function; calculating the fitness value of the current population, and updating related parameters in the dolio-sea squirt and moth fire suppression strategy; setting related parameters of a sine and cosine optimization algorithm, and updating the population by adopting a sine and cosine optimization algorithm iterative formula; updating the populations obtained through the sine and cosine optimization algorithm through a doliola scabra, moth fire suppression and reverse learning strategy in sequence so as to obtain three populations; selecting a next generation of population through greedy selection; and if the maximum number of iterationsis reached, ending the loop and outputting the optimal solution, otherwise, continuing the iteration until the iterative computation is ended. According to the invention, the gene characteristics which contribute most to categories can be screened out from the genes more accurately and more efficiently, and the detection cost is reduced.

Allocation of CSI-RS or CSI pilot sequences among cells in a cooperative manner to reduce pilot inter-cell interference. Sequences in each cell occupy the same time slot, multiple subcarrier frequencies, and are orthogonal in time via properly chosen cyclic shifts. Sequences in multiple cells are chosen from a pool of non-orthogonal yet distinguishable sequences through their root indices. Exchanging root indices among cells allows a given cell to reconstruct sequences used in neighboring cells and to estimate interfering channels as the number of channel taps is usually limited, thus mitigating pilot contamination. Furthermore, a greedy selection algorithm to find combinations of sequences that further reduce the channel estimation mean-square-error is proposed.

A group path planning method based on the variable dimension ABC algorithm, characterized in that: the path planning problem of the group system is taken as the research object, combined with the ability of the ABC algorithm to solve the multi-variable function global optimization speed, and the path of the group system The characteristics of the planning problem are improved. In the initial stage of the path network, the endpoint heuristic method is used to generate the initial solution set in the feasible path solution space, and then the path network is optimized by the MDABC algorithm. A new ABC model is proposed in the optimization process. , it can adaptively adjust the fitness of the solution according to the dimension change of the solution in the solution space, select the optimal solution through the greedy selection mechanism, and obtain the waypoints of the path network that satisfy the robot motion constraints. The multiple robot obtains a safe path network between the task set by connecting the waypoints. That is, it maintains the strong global search ability of the ABC algorithm, and at the same time solves the problem of the algorithm falling into local optimum and poor robustness when facing the path planning problem of multiple robot clusters with multi-constraints and multi-tasks.

The invention discloses an uplink transmission method for a periodic URLLC (Uniform Resource Location Link Control) service. The uplink transmission method comprises the following steps: 1) constructing a URLLC uplink transmission model; 2) constructing a URLLC service uplink transmission QoS guarantee mechanism based on a finite coding length theory; 3) by formulating a reasonable sub-channel-power joint distribution strategy, while the uplink transmission QoS of all devices in the system is guaranteed, the required total transmitting power is minimized, so that an edge adaptive optimizationproblem is constructed; and 4) respectively solving the edge adaptive optimization problem in the step 3) through a Kuhn-Munkres algorithm based on a weighted bipartite graph theory or a low-complexity algorithm based on greedy selection, and obtaining an optimal sub-channel-power distribution strategy, and completing uplink transmission of the periodic URLLC service according to the optimal sub-channel-power distribution strategy. The method can be suitable for uplink transmission of the periodic URLLC service.

The invention provides a polarity searching method of a fixed-polarity Reed-Muller logic circuit. The best polarity of an FPRM logic circuit is searched by using the new binary differential evolution algorithm, compared with a polarity searching method of the FPRM logic circuit based on the genetic algorithm, the capacity of local optimum and the capacity of premature convergence are avoided, and the convergence rate and the polarity searching efficiency are improved. The method comprises the following steps that firstly, a Boolean logic circuit is read; secondly, an evolution parameter is input; thirdly, an initial population is generated randomly, wherein the polarity is encoded into a binary individual; fourthly, the improved binary stochastic mutation operation is performed; fifthly, the binomial crossover operation is performed; an FPRM expression of the target individual and the FPRM expression of a test individual of the target individual are obtained; seventhly, the fitness value of the target individual and the fitness value of the test individual of the target individual are calculated; eighthly, the greedy selection operation and an elitism selection strategy are performed; ninthly, if the current evolution algebra is smaller than the maximum evolution algebra, the fourth to eighth steps are executed in sequence, and otherwise, the best polarity is output.

The invention discloses a satellite Internet of Things routing strategy based on Q-Learning. Aiming at a satellite Internet of Things routing problem in a complex environment, a satellite Internet ofThings topological structure and node state dynamic changes are considered, the whole satellite Internet of Things is regarded as a reinforcement learning environment, and satellite nodes and ground nodes are regarded as intelligent agents., A method for implementing the strategy comprises the following stepsL firstly, initializingand satellite Internet of Things parameters are initialized firstly; secondly, maintaining a Q value table by each node, and learning the Q value table by utilizing a Q value updating formula according to the hop count, the distance, the direction and the buffer areaoccupancy rate of the satellite nodes; and finally, forwarding the data packet according to a greedy selection strategy through a Q value table obtained by learning. Moreover, the reward value is improved by considering the hop count of the satellite nodes, and the discount factor is improved by considering the distance and direction of the satellite nodes and the occupancy rate of the buffer area, so that the Q value is optimized, and the purpose of efficient routing of the satellite Internet of Things is achieved. Therefore, the satellite Internet of Things routing strategy has good conversion and application prospects in the fields of aviation, spaceflight, social economy and the like.

The invention provides a breast cancer image feature selection method based on an improved sine and cosine optimization algorithm, and the method comprises the steps: extracting feature data of breastcancer image features to acquire a training sample set, and initializing a population; designing a support vector machine classifier according to the training sample set, and performing classification; calculating the fitness value of the current population, and updating related parameters in the policies of the sea squirt and the grey wolf; setting related parameters of a sine and cosine optimization algorithm, and obtaining a population updated through the sine and cosine optimization algorithm; updating the obtained populations updated by the sine and cosine optimization algorithm througha sea squirt flight strategy, a grey wolf flight strategy and a Levy flight strategy to obtain three populations; screening out an optimal population through greedy selection; and if the termination condition is met, ending and outputting the optimal solution, otherwise, continuing iteration until the iterative computation is ended. By implementing the method, the problems of falling into a localoptimal solution, low convergence rate and the like of a sine and cosine optimization algorithm can be solved, and classification and prediction of breast cancer images are realized.

The invention provides a novel optimistic inquiry method applicable to a Chord-like environment. The method comprises the following steps of: firstly, by utilizing a manner that all nodes on a route path share the feedback of a start node, collecting feedback messages and form a training set; utilizing a Beta estimation mechanism to carry out trust measurement on a network node historical route behavior; meanwhile, aiming at a short-period online behavior of chord network nodes, determining a minimum transmitting time threshold value through setting the confidence coefficient; and ignoring a provided reputation calculation result by a system aiming to a reputation value provided by the node with the transmitting time which is smaller than the threshold value, so as to guarantee the reliability of the formed reputation value. On the aspect of a selection manner of the transmitting node, a manner of combining a greedy policy and a reputation policy is adopted; firstly, the greedy policy is used for selecting the next jump; if the inquiry is failed, a principle of reputation selection is used for gradually replacing greedy selection when starting from a source starting node; and the method can realize the reasonable multi-path route and considers both the inquiry efficiency and the inquiry success rate.

The invention discloses a trust-based mobile edge user task scheduling method, which selects a reliable user to execute a task by performing trust evaluation on a mobile edge user so as to realize optimization of task scheduling. The method is divided into two parts. The method comprises the following steps: 1, recruiting mobile edge users with high trust degree and low cost to execute tasks based on a greedy selection algorithm considering trust degree, thereby obtaining a credible evaluation result, accurately distinguishing the reliability of Internet of Things equipment in a network, and collecting data generated by reliable equipment; and 2, comprehensively applying an aggregation-based trust reasoning mechanism and an unmanned aerial vehicle-assisted active trust acquisition mechanism to perform trust evaluation and reasoning on the mobile edge users. An unmanned aerial vehicle is introduced to actively verify data uploaded by a mobile edge user, so that the accuracy of trust evaluation is improved. The trust-based task scheduling optimization method provided by the invention can effectively improve the recruitment performance of the cloud and reduce the recruitment cost.

The invention discloses an antenna selection method based on a sub-mode function for a large-scale multiple input multiple output (MIMO) scenario, which solves the technical problems that the complexity is high due to too many antennas and the channel capacity of the conventional antenna selection method is not high enough. The method comprises the steps of: defining antenna selection parameters;establishing an antenna selection model of the sub-mode function; determining criteria for greedy selection; approximating the model with a greedy selection algorithm; and completing the antenna selection. The method intuitively transforms the antenna selection into a discrete problem of the sub-mode function from the idea of set selection, and solves a mathematical model to complete the antenna selection. A greedy selection approximation algorithm for solving the discrete problem of the sub-mode function is designed, with the approximate ratio (1-1 / e), which simplifies the calculation and reduces the complexity. The implementation complexity of the method is low, the channel capacity performance of the system is improved under the same condition, the performance is guaranteed in any practical application, and the method can be used for fifth-generation mobile communication large-scale MIMO systems and cell mobile communication systems.