32results about How to "Speed up evolution" patented technology

The invention discloses a multi-target evolutionary fuzzy rule classification method based on decomposition, which mainly solves the problem of poor classification effect of an existing classification method on unbalanced data. The multi-target evolutionary fuzzy rule classification method comprises the steps of: obtaining a training data set and a test data set; normalizing and dividing the training data set into a majority class and a minority class; initializing an ignoring probability, a fuzzy partition number and a membership degree function; initializing an original group, and determining weight by adopting a fuzzy rule weight formula with a weighting factor; determining stopping criteria for iteration, iteration times, a step size and an ideal point; dividing direction vectors according to groups; performing evolutionary operation on the original group, and updating the original group by adopting a Chebyshev update mode until the criteria for iteration is stopped; obtaining classification results of the test data set; then projecting to obtain AUCH and output. The multi-target evolutionary fuzzy rule classification method has the advantages of high operating speed and good classification effect and can be applied in the technical fields of tumor detection, error detection, credit card fraud detection, spam messages recognition and the like.

The invention relates to an in-vitro directed coevolution method for modifying L-phenylalanine gene engineering strains, which is realized in a way that: using genes aroG and pheA on an L-phenylalanine gene engineering strain constructed in the room as a whole; carrying out the in-vitro directed coevolution modification by using an error-prone PCR technology and a recombinant DNA technology; and screening to obtain a mutant strain, of which the yield of L-phenylalanine is increased by 114%. In the invention, by modifying the gene formed by coupling and connecting aroG and pheA in series, the key genes aroG and pheA in the metaboly process of the L-phenylalanine are used as a whole to carry out the directed modification, thereby obtaining a new metabolic balance, so that the expressed enzyme has high-efficiency catalytic activity and can resist the feedback inhibition of the L-phenylalanine, thereby obtaining the new modified high-yield L-phenylalanine gene engineering strain by screening. The method can provide an example for modifying the acid production rate of any amino acid gene engineering strain.

The invention discloses an intelligent reservoir optimal operation method. According to the intelligent reservoir optimal operation method, based on the analysis on the defects of a traditional simulated annealing genetic algorithm in the aspect of reservoir optimal operation, improvement is conducted on the algorithm by introducing the ecological niche technology, adopting the self-adaptive crossover and mutation strategy and adopting the optimization saving strategy in the selection process, and a reservoir optimal operation nonlinear mathematical model with the maximum generating capacity as the target is solved according to the specific condition of a reservoir.

The invention relates to a method for improving the acid production rate of L-phenylalanine gene engineering bacteria. The method comprises the following steps of: directionally mutating recombinant expression plasmids of L-phenylalanine gene engineering bacteria constructed in the laboratory, which are taken as a whole; in the process of mutating the expression plasmids with an alkylation agent EMS, controlling the mutation rate by the length of the duration of the reaction of the EMS and the plasmids to obtain the optimal mutation time under the same mutation conditions, and then electrically converting the expression plasmids after the optimal mutation time of the EMS mutation into host bacteria to construct a strain mutation library; and screening to obtain 4 good mutant strains the yields of L-phenylalanine of which are improved in different degrees. The method ensures the reconstruction of acid-producing strains of L-phenylalanine gene engineering bacteria which are easy to be large-scale fermented and industrialized, and facilitates the reconstruction.

The invention belongs to the field of power distribution network planning and particularly relates to a reactive configuration optimization method of a power distribution network under a microgrid connection condition, and a reactive compensation equipment optimal configuration method of coordinated planning of a switchable parallel capacitor bank (C) and a static var generator (SVG) (the structure is shown in a figure) is proposed. Firstly, the microgrid connection point load characteristics are considered, on the basis, a double-layer optimal planning model for coordinated configuration of dynamic and static reactive equipment is established, the upper layer model takes the sum of the life cycle cost (LCC) of the equipment and the grid operation cost as a target function, and the configuration mode of the reactive compensation equipment is optimized; the lower layer model takes the minimum operation cost as a target function, the operation mode of the capacitor bank is optimized, andmodel solution is carried out by adopting an improved genetic algorithm. Finally, the validity of the model is verified with a modified IEEE33 node as an example. The simulation result shows that through the planning method, the power quality can be obviously improved, and the network loss can be reduced; and compared with the traditional method, the coordinated planning method can be more suitable for the active power distribution network.

The present invention generally relates to a controlled fluidic device to develop spatially complex environments to enhance the rate of evolution in cell populations. The method further provides an enhanced understanding in the emergence, for example, drug resistance during cancer chemotherapy.

The invention relates to a wireless sensor network and a node task distribution method thereof. The wireless sensor network comprises a gateway node and multiple common nodes. The wireless sensor network is characterized in that the gateway node and the multiple common nodes are formed through linkage in a wireless multi-hop mode, the gateway node is provided with power supply, and the common nodes are provided with no power supply, are randomly arranged and are moved no longer once the common nodes are arranged well. The gateway node in the wireless sensor network allocates interdependent subtasks based on a genetic algorithm, an integral incentive mechanism is introduced to an allocation algorithm, grading is conducted on the common nodes according to node integrals, accordingly loads of all nodes are balanced, and the life cycle of a network is prolonged. Under the situation that the nodes are unstable, uncompleted tasks on unstable points can be migrated to other appropriate nodes, task completion within time limit is ensured, and task allocation efficiency and completion quality are improved.

The invention relates to the technical field of AUV path optimization, and provides an AUV energy optimization path searching method based on distance evolution N-PSO. The AUV energy optimization pathsearching method comprises the steps of: firstly, constructing an underwater environment model and an AUV two-dimensional motion model; secondly, generating an initial path of particles in a particleswarm randomly based on N-PSO, updating a global optimal solution and an individual optimal solution according to a particle penalty function value in the k-th iteration, and constructing a distanceevolution factor and an evaluation state Evo_statek according to the average distance among the particles; when Evo_statek is equal to 1, outputting an energy optimal path and its energy consumption value if k>=K, otherwise, updating a particle velocity and particle positions and carrying out next iteration; and when Evo_statek is not equal to 1, randomly perturbing the particles and updating a global optimal solution and an individual optimal solution when penalty function values of the particles become smaller or iter>=ITER after perturbation, otherwise, carrying out next iteration. The AUVenergy optimization path searching method can optimize the AUV path from the perspective of energy optimization, and has the advantages of high optimization efficiency, good robustness, more stable optimization result and easy implementation.

The invention discloses a reuse method of test cases between similar programs and an implementation system thereof. The reuse method comprises the following steps: based on a similarity judgment process of a program to be tested, constructing a keyword flow graph, comparing whether keywords in keyword flow graph nodes are the same or not, and forming a public keyword flow graph subgraph by the nodes with the same keywords; after the construction of the keyword flow graph of the program and the maximum public subgraph of the keyword flow graph is completed, comparing the similarity of the program to be tested by using a maximum public subgraph distance method, and reusing the program of which the similarity degree reaches a preset value for the test case. According to the method, the existing test cases are shared with the similar programs, the reused test cases are generated by adopting a genetic algorithm, the test cases generated by the similar programs are introduced into the population evolution process, newly generated population individuals learn to the test cases to accelerate the evolution speed, and reuse of the test cases is completed.

Provided is a method for acquiring a residual static correction value. The method comprises the steps that common midpoint way set data after dynamic correction are read in, and statistics is carried out on the number of shot points in the data and the number of demodulator probes in the data; according to the number of the shot points, the number of the demodulator probes and a preset residual static correction value search range, an initial solution set is generated; a Poisson disk sampling algorithm is utilized for obtaining a sampling solution set according to the initial solution set, and the sampling solution set is mapped to a genetic space through real number coding to form an initial population; the genetic manipulation is carried out on the initial population, and an individual meeting the preset conditions is acquired as the residual static correction value. According to the method, the evolution speed and effect of the solving process of the residual static correction value can be improved.

The invention discloses a multi-energy short-term economic dispatching method and system based on differential-gradient evolution and belongs to the field of multi-energy optimization dispatching. Themethod comprises steps that for power grids including a wind power plant, a hydropower station and a thermal power unit, a model is built at the minimum coal-fired cost of the thermal power unit on the premise of considering the valve point effect, and model constraints are set; the wind speed is utilized to calculate output of the wind power plant, and output of the hydropower station is calculated through utilizing the inbound flow; output of the thermal power unit is utilized as a decision variable, based on the output of the wind power plant and the output of the hydropower station, the model constraints are utilized to compress the decision space; iterative evolution is performed through utilizing a differential-gradient algorithm in the decision space to obtain the best output of the thermal power unit, and the output of the wind power plant, the output of the hydropower station and the optimal output of the thermal power unit are combined to obtain a scheduling plan of the power grid. The method is advantaged in that the method is not easy to fall into local optimum and can solve a technical problem of dynamic economic dispatching of a multi-energy mutual aid coordination system.

The invention provides a multi-parameter optimization method. With the multi-parameter optimization method adopted, learning efficiency can be improved, and time complexity and space complexity can be decreased. The method includes the following steps that: the position and the number of iterations of each particle in a particle population are initialized, each dimension of the particle is corresponding to a parameter to be solved; the historical optimal position of each particle and the optimal position of the population are updated according to the calculated fitness value of each particle to an objective function, and an orthogonal algorithm is utilized to perform orthogonal calculation on the historical optimal position of each particle and the optimal position of the population, so that the learning vector of each particle can be obtained, the locations of particles of a next generation are determined according to the position and learning vector of each particle; and if reverse learning is performed, each particle is reversed, so that reverse particles can be obtained, if the fitness value of the reverse particles to the objective function is larger than the fitness value of the original particles to the objective function, the reverse particles are adopted to replace the original particles. The multi-parameter optimization method of the invention is applicable to the industrial parameter optimization technical field.

The invention discloses a method for treating spartina alterniflora by using habitat restoration ecological succession measures. The method utilizes the excellent growth characteristics of pioneer mangrove forest species, the growth speed of the pioneer mangrove forest species exceeds that of spartina alterniflora in a short time and the pioneer mangrove forest species is quickly closed, so that the growth of the spartina alterniflora is inhibited, the growth and spreading of the spartina alterniflora are controlled, an arbor, shrub and vine replacement plant multi-layer community is gradually formed, a coastal zone three-dimensional composite ecological system is finally constructed, and therefore the purpose of treating the spartina alterniflora is achieved; and the method overcomes the defects of physical, chemical and biological prevention and control of spartina alterniflora in the prior art, avoids pollution caused by chemical control of foreign species and ecological risks possibly caused by control of natural enemies, saves time and labor, is good in prevention and control effect, achieves large economic and ecological benefits, and is one of the best ways for controlling harm of foreign invasive plants.

The invention provides a maximum complete subgraph-based embedded system register allocation method, which mainly solves the problems that a heuristic algorithm has poor allocation effect and overflow cost is too high due to no consideration of the overflow cost in a crossover operator part in an evolutionary algorithm. The method is implemented by the following steps of: (1) taking a complement from an intermediate variable mutual interference graph to obtain the complement G; (2) randomly dividing all nodes into two classes, respectively putting the nodes into a set A or a set B, and completing initialization of population; (3) crossing an individual in the population by using an overflow cost-based maximum complete subgraph crossover operator SC-MCX to generate a subgeneration individual; and (4) optimizing the subgeneration individual by using a local search operator LSP, replacing an individual having the maximum fitness function value in parent individuals by using the optimized subgeneration individual, and continuously participating in population evolution. By the method, the population evolution speed is improved, the overflow cost and the overflow variable number of the individual are reduced, and the method can be used for embedded system register allocation.

The invention discloses a supply chain financial credit analysis method based on a multi-layer genetic method under the background of big data. The invention designs a data classification method to perform fuzzy processing on it, effectively speeding up the computer's analysis and calculation of data; at the same time , in the coding stage, this embodiment divides the supply chain data into three layers according to the characteristics, and each layer designs a corresponding crossover method. As the genetic segment at the bottom layer, this embodiment designs random intersection points to meet the diversity of the population; as the chromosome at the middle layer, it plays a vital role in finding the local optimal solution, so this embodiment uses the golden section method to design Fixed intersection points, while avoiding the interaction between different types of data; as the top layer of the nucleus, the purpose of the intersection is to gather chromosomes with high fitness into one chromosome, speed up the evolution of the nucleus, and improve the convergence speed of the method.

The present invention relates to the technical field of AUV path optimization, and provides an AUV energy optimization path search method based on distance evolution N-PSO. First, an underwater environment model and an AUV two-dimensional motion model are constructed; Generate the initial path, in the kth iteration, update the global optimal solution and the individual optimal solution according to the value of the particle penalty function, and construct the distance evolution factor and evolution state Evo_state according to the average distance between particles k ;Evo_state k =1, if k≥K, then output the energy optimal path and its energy consumption value, otherwise, update the particle velocity and position, and proceed to the next iteration; Evo_state k When ≠1, the particle is randomly disturbed. When the penalty function value of the particle becomes smaller after the disturbance or iter≥ITER, the global optimal solution and the individual optimal solution are updated, otherwise, the next disturbance is performed. The invention can optimize the AUV path from the perspective of energy optimization, has high optimization efficiency, good robustness, more stable optimization results, and is easy to implement.

The invention discloses a fluid filter and a filtering method. The fluid filter comprises a shell into which a to-be-filtered fluid is introduced; a rotating impeller of which a plurality of blades drive the fluid to rotate and enable the fluid to generate centrifugal force; and a plurality of filter blocks which are fixed on the inner wall of the shell. A high-pressure cavity is formed at the inlet of each filter block when the blades rotate. The filter blocks are used for separating fluid into first-stage first liquid and first-stage second liquid, the first-stage first liquid flows out of the shell, the first-stage second liquid enters the next adjacent filter block and continues to be separated into second-stage first liquid and second-stage second liquid, the second-stage first liquid flows out of the shell, the second-stage second liquid enters the next adjacent filter block and continues to be separate into third-stage first liquid and third-stage second liquid, and the process is repeated. Fluid centrifugal force is generated through circular motion, so that fluid continuously rotates to enter the multi-stage filter blocks to be separated, the multi-stage separation and filtration effect is good, and the purity is higher.

The present invention provides a method for obtaining the expected formation of the unmanned aerial vehicle cooperatively tracking the ground target under the implicit communication, a geometric structure and a topological structure can be combined for analysis, and the obtained target formation information can well meet expectation. According to the method for obtaining the expected formation of the unmanned aerial vehicle, a geometric structure and a topological structure are combined for analysis, and accurate target formation information is obtained. Specifically, for a ground target tracking task of an unmanned aerial vehicle formation, constraints such as a limited detection range of a sensor and a shortest distance (collision prevention) between unmanned aerial vehicles are considered, and an Euclidean distance and a relative sight angle between the unmanned aerial vehicles are taken as decision variables; and a formation optimization problem is established by taking the maximum topological structure survivability and tracking task stability (determined by a geometric structure) of the formation as an objective function, and an expected formation is obtained by applying an optimization problem solving algorithm.

The invention discloses a square inner cavity nozzle with a noise reduction function. The nozzle structurally comprises a base, a nozzle head, a nozzle core, a bamboo joint pipe and a liquid conveying hose. The nozzle head is coaxially installed at the front end of the base, the bamboo joint pipe is installed at the rear end of the base, the nozzle core is installed in inner cavities of the base and the nozzle head, and the rear end of the nozzle core is connected with a fluid source through the infusion hose. The nozzle head inner cavity is a square-column-shaped nozzle inner cavity, and the local section of the nozzle core is square. Extra equipment does not need to be added, evolution of a shear layer in the nozzle is accelerated by changing the shape of the nozzle core and the section shape of the inner cavity of the nozzle, the complete structure of a vortex is destroyed, the purpose of reducing noise is achieved, the manufacturing process is slightly changed, and the nozzle has the advantages of being simple and economical.

The invention relates to the technical field of image processing, and discloses a machine vision-based method for classifying and identifying formed elements in feces microscopy images. Firstly, the level set method is used to segment the feces microscopic image, and then for the shape contour of the formed components obtained by segmentation, the primary classification based on morphological features and the verification and recognition based on HOG+VSM training classifier are sequentially performed, and finally quickly Obtain high-precision classification and identification results of formed components, so that patients can receive medical treatment in time. In addition, this method also has the advantages of high image segmentation accuracy, fast processing speed and good user experience, which is convenient for practical application and promotion.

The present invention relates to a chaotic genetic BP neural network image segmentation method based on Arnold transformation, the method comprises adopting a chaotic genetic algorithm to optimize a BP neural network, and utilizing a trained BP neural network to perform image segmentation; the specific process of the chaotic genetic algorithm is: ① Initialize the population: Use the chaotic map to generate two populations x and y, use the small population x as the initial population, and the large population y as a backup; ② calculate the individual fitness value in the initial population x; set the individual fitness value in the initial population x Finally, replace the set number of individuals with individuals in the large population y, and calculate the fitness value of the replaced individual; ③ according to the calculated individual fitness value, perform selection, crossover and chaotic mutation operations on the individuals in the initial population x The algorithm terminates until the maximum number of evolutions is reached or the maximum fitness does not change. The invention can effectively ensure the ergodicity of the population evolution process, accelerate the neural network training process, and enhance the image segmentation effect.

The invention discloses a level set image segmentation method and a system thereof based on regional information and edge information. The level set image segmentation method comprises the steps of: constructing an energy functional and initializing a level set function according to the regional information and edge information of an image, so as to obtain an initial level set function; adopting a gradient descent flow method to obtain a level set function evolution equation; and carrying out level set function evolution processing at a coarse segmentation phase and a smooth phase based on the energy functional, and carrying out rough segmentation and accurate positioning of the image to obtain the segmented image. The system comprises an initialization module, a level set function evolution module, a color variance calculation module and a segmented result output module. The level set image segmentation method and the system thereof reduce the number of iterations of the level set function, and improve the segmentation and positioning ability of the level set on a color image.

*The invention discloses a machine multi-rotating-speed workshop energy-saving scheduling method based on an improved SPEA2 algorithm. The method comprises the following steps: firstly, constructing a flexible job shop energy-saving scheduling problem model which comprises flexible job shop energy-saving scheduling problem description and model construction; natural number-based three-section coding is adopted, and the three-section coding comprises a process code, an equipment code and a speed code; fitness assignment is carried out, a non-dominated solution set is constructed through rapid non-dominated sorting, then environment selection is carried out, and an optimal compromise solution is selected by using a weighting method; terminating the judgment, and if t is greater than or equal to tmax, outputting an optimal compromise solution X *; on the contrary, if t is smaller than tmax, genetic selection is executed; dividing populations A, B, C and D through indexes such as a special crowding degree distance and an average fitness value; and selecting individuals in the populations A, B, C and D by using a simulated annealing selection mechanism, and respectively carrying out reverse learning, spiral position updating, Gaussian variation or random disturbance and multi-bit transformation operation. According to the method, the problem of over-high energy consumption in the workshop production process in the prior art is solved.

The invention discloses a single-target flexible job shop energy-saving scheduling method based on an improved grey wolf algorithm, and the method improves the grey wolf algorithm, employs two-segment coding, and achieves the conversion between a continuous individual position and a discrete scheduling solution. A population initialization method is designed, and the quality and diversity of the population are ensured; a dual-mode parallel search mode is used, and the global search capability and the local search capability of the algorithm are balanced in the search process; the improved tracking and searching operation is introduced, the problem solving precision is improved, combined variation is adopted for searching, the evolution speed is increased, and a local optimal solution is broken through. In addition, a better individual in the two sub-individuals obtained by the tracking operation is selected as a new individual according to the fitness value relationship.

The invention discloses a method for automatically segmenting a face in a digital photo. The method comprises the following steps of: (1) in order to avoid local optimization, combining an edge tension and skin color information with global energy, and designing a geometric active contour model based on unified energy functional; (2) designing an improvement method based on a single Gaussian model, so that an initial position approximate to actual face outer contour is estimated; (3) according to the neighborhood information of an evolved curve, endowing the outward tension of edge points, sothe curve can overcome the interference of facial features and facial decorations and can be guided to be evolved to the direction of the outer contour; and (4) as skin color is the most important facial feature, designing the skin color energy and further improving the robustness of the single Gaussian model. By the method, the interference of factors such as attitude, illumination and complicated background can be overcome; an accurate segmentation effect is obtained; and the method can be widely applied to face image pre-processing such as face identification, games and automatic focusing of digital cameras.

The invention discloses a multi-objective evolutionary fuzzy rule classification method based on decomposition, which mainly solves the problem that the existing classification method has poor classification effect on unbalanced data. The implementation steps are: obtain training data set and test data set; normalize and divide the training data set into majority class and minority class; initialize ignore probability, fuzzy partition number and membership function; initialize the original population, adopt weighting factor The fuzzy rule weight formula determines the weight; determines the iteration termination condition, the number of iterations, the step size, and the ideal point; divides the direction vector according to the population; performs the evolution operation on the original population and uses the Chebyshev update method to update the original population; until the iteration termination condition; obtains The classification result of the test data set; re-projection to get AUCH and output. The invention has fast operation speed and good classification effect. It can be applied to technical fields such as tumor detection, error detection, credit card fraud detection, and spam short message identification.

The invention discloses a space debris orbit rapid evolution method, and belongs to the field of spacecraft orbit prediction. The states of a plurality of fragments are described as a nominal state and a corresponding state deviation distribution form through a Taylor polynomial, and polynomial integration is used for replacing a large number of repeated fragment orbital integral operations. A large amount of fragment evolution is changed into a numerical value with an extremely high speed to be brought into the process, the overall calculation efficiency is greatly reduced, and the acceleration effect of several orders of magnitudes can be achieved. By adopting high-order Taylor expansion, the operation precision of the technical scheme provided by the invention can theoretically reach the precision the same as that of the traditional Monte Carlo targeting method. By balancing the calculation precision and the calculation efficiency, the error of the five-order approximation method issmaller than 0.01 m, and the situation that large precision is sacrificed due to accelerated operation is effectively avoided. According to the method, a large number of space debris orbits can be rapidly evolved, the orbital evolution speed of a large number of debris is greatly increased, and the orbital precision in the whole evolution process can be guaranteed.

The invention discloses an evolvable pulse neural network construction method and device. The method comprises the steps of initializing an evolvable pulse neural network; training the evolvable pulseneural network by using training samples, calculating the corresponding moment of generating a pulse in a potential time region after the samples are subjected to the sudden contact, and selecting network evolution strategies based on the pulse generation time to evolve the neural network, wherein the network evolution strategies comprise an output neuron adding strategy, an input pulse sequencecanceling training strategy and a weight parameter updating strategy. By means of selection of three evolutionary strategies, unnecessary updating of network weights can be avoided, and network evolution speed is increased.

The invention discloses a bionic hybrid evolution method based on an octopus learning and memory system, which comprises the following steps: firstly, setting individual codes and an initialized population, and then constructing a learning and memory network: using two networks S and V, using the S as a sensing network for simulating a first fan-out synaptic layer, using the V network as a memory network corresponding to a fan-in synaptic layer of a vertical leaf structure, and using the memory network as a second fan-out synaptic layer of a vertical leaf structure for simulating a second fan-out synaptic layer; selecting individuals according to values in the learning memory network, and then carrying out individual crossing; finally, performing individual mutation operations according to the high and low mutation probability. According to the method, the learning memory network is introduced to select the individuals, the individuals which are determined to be the local optimal region are removed, the situation that the population individuals fall into the searched region can be effectively inhibited, the population evolution speed can be increased, and the population can effectively jump out of the local optimal solution.