40 results about "Greedy selection" patented technology

The invention provides a cue word optimization method and system based on an approximate sub-module function and continuous learning, and relates to the technical field of artificial intelligence multi-mode perception.The method comprises the steps that a candidate cue word set is constructed, a combined objective function based on the property of the approximate sub-module function is designed, and a candidate cue word set is constructed; solving the combined objective function by adopting a greedy selection algorithm combining random disturbance, multi-round iteration and a task self-adaptive mechanism, realizing optimal selection of a candidate cue word set, gradually selecting a cue word with the maximum gain from the candidate cue word set, adding the cue word into an optimal subset, and when a new cue word is selected, selecting the cue word with the maximum gain into the optimal subset. And if the target cue words are selected, carrying out local optimization once, after all the target cue words are selected, carrying out joint optimization on all the cue words in a continuous space by adopting an alternate optimization strategy, and carrying out continuous iteration until the optimized cue words are obtained. According to the method and the device, efficient self-adaptive updating of the cue words of the language model is realized, so that the generalization performance and robustness of the model in zero-sample, few-sample and concept drift scenes are improved.

The invention provides an unmanned tractor path planning method based on an improved PSO fusion DWA optimization algorithm. The method comprises the steps that a grid map is initialized; forming a preliminary global path by using an improved PSO algorithm; performing local refined optimization on the initial global path through single-dimensional neighborhood disturbance and greedy selection; discrete path serialization, redundant node elimination and smoothing processing are carried out on the initial global path through a greedy optimization method; a path point sequence of a final optimized global path is completely reserved as a local target point of a DWA algorithm, local path planning and real-time obstacle avoidance are guided, speed sampling and trajectory prediction are performed based on a kinematic model, speed constraint and the local target point, and an optimal prediction trajectory and speed space are selected through an improved evaluation function in the DWA algorithm. The method effectively solves the problems that an existing planning method is poor in smoothness, long in planning time, multiple in inflection points, discontinuous in path curvature and incapable of achieving global optimum.

The invention discloses an unmanned aerial vehicle area coverage flight path planning method and system based on graph segmentation, and relates to the technical field of unmanned aerial vehicle path planning. According to the method, a coverage point pool with high quality, comprehensive coverage and robustness is constructed for subsequent greedy selection through a strategy of adaptively generating candidate points; then, a greedy selection mechanism based on effective scores is adopted, the current optimal coverage point is accurately selected in each round of iteration, the new coverage area can be increased to the maximum extent, the overlapping area can be effectively controlled, and finally the whole target area is completely covered with the minimum number of circles. In the aspect of a path planning algorithm, a hybrid initialization mode of a greedy algorithm and a random generation strategy is innovatively combined, and the quality of an initial solution and population diversity are ingeniously balanced; meanwhile, in each generation of evolution of the genetic algorithm, a 2-opt local search strategy is embedded into a new non-elite individual, so that the convergence speed of the algorithm is increased, and the path optimization efficiency is remarkably improved. In conclusion, more efficient and more reliable planning of the unmanned aerial vehicle area coverage flight path can be realized.

The invention relates to the field of structural health monitoring, in particular to a three-axis sensor number and position optimization method for structural response reconstruction. Aiming at the information coupling characteristics of the three-axis sensor in all directions, constructing a redundancy evaluation index based on the difference degree of a three-axis information matrix, screening the minimum initial measuring point set, removing measuring points with repeated information contribution, and forming a non-redundancy candidate measuring point set; based on a non-redundant candidate measuring point set, a forward greedy selection algorithm is adopted to carry out iterative optimization on the number and positions of sensors, unmeasured redundant measuring points enabling marginal gains to be maximum are selected to be added into an optimization set, it is ensured that redundancy is effectively removed in the initial stage of iteration, the redundancy is moderately relaxed in the later stage, and dependence on prior knowledge of a target structure is reduced; an objective function value changing along with the number of the sensors is used as an uncertainty curve, and the optimal critical value of the number of the sensors is determined by combining an inflection point criterion and a platform criterion.

The invention discloses an image segmentation optimization method, system, equipment and medium, and relates to the technical field of computer optimization algorithms, and the method comprises the steps: constructing a two-dimensional histogram for an original gray level image, employing a two-dimensional Renyi entropy as a target function for evaluating the segmentation quality of the two-dimensional histogram, finding a group of optimal threshold combinations, and carrying out the segmentation of the two-dimensional histogram through the optimal threshold combinations; the problem of searching the optimal threshold value combination is converted into an optimization problem to be solved, the found optimal threshold value combination is used for segmenting the original image, and a final segmentation result image is generated; when an optimization problem is solved, operating individuals in the population by adopting a differential variation strategy so as to generate a first group of candidate solutions, and executing a Rime optimization mechanism and a randomized step size covariance matrix adaptive mechanism in parallel on each individual in the population so as to generate a second group of candidate solutions; and updating the population according to a greedy selection mechanism. The algorithm premature convergence can be effectively prevented, and the method can be applied to complex optimization tasks such as image segmentation and engineering design.

The invention discloses a multi-unmanned aerial vehicle path planning method and system based on a quantum song eagle mechanism and a storage medium, relates to the field of unmanned aerial vehicle path planning, and aims to solve the problem that the performance of an existing method is remarkably reduced and even fails in a complex scene. Comprising the steps of 1, establishing a complex environment and threat source model; 2, setting a cost function in a path planning process of the unmanned aerial vehicle; step 3, initializing a quantum Song eagle algorithm; 4, calculating a fitness value and an optimal quantum position of each song eagle; 5, in an exploration stage, using Cauchy inverse cumulative distribution variation to enhance population diversity and using a quantum revolving door to evolve quantum song eagle positions; 6, in a development stage, a dynamic boundary strategy is used to avoid falling into a local optimal solution, and a quantum revolving door is used to evolve a quantum song eagle position; 7, determining the position of the next generation of quantum eagle by adopting a greedy selection strategy; and 8, repeatedly executing the steps 5-7, and outputting a global optimal quantum song eagle position to obtain a path planning result.

The invention relates to a low-overhead moving target defense method and device for false data injection attacks, and the method comprises the steps: constructing a topological structure of a power grid bus and branches based on actual intelligent power grid data, and associating the branches with an initial admittance value; establishing an initial measurement matrix based on the initial admittance value and the topological structure of the power grid bus and the branches, and constructing a moving target defense strategy based on the initial measurement matrix; based on an invalid branch identification criterion and a topological structure of a power grid bus and a branch, determining an invalid branch in the power grid; based on invalid branches and a greedy selection mechanism, traversing the bus to screen valid branches, and obtaining a modified branch set; and changing admittance values of branches in the branch set based on a moving target defense strategy, changing a measurement matrix, and realizing low-overhead moving target defense. Compared with the prior art, on the premise that the high detection probability is maintained, the number of branches needing to modify admittance and the calculation time are remarkably reduced, and therefore the system defense cost is reduced.

The present application relates to the technical field of large model, and more particularly to a data set semantic consistency generation and quality stratification method for large model training, which comprises the following steps: step one: performing semantic coding on candidate samples containing synthetic samples and original samples to obtain unit direction vectors, determining radial coordinates according to the semantic sparsity of the nearest neighbor original samples, and dividing radial shell layers in the Poincare ball hyperbolic space; step two: establishing a support cone for the original sample, determining the synthetic samples falling into the inner support cone and anchored with the shell layer as the semantic consistent samples together with the original sample, and determining the quality level according to the radial shell layer number; step three: constructing a core matrix in each quality level to perform a determinant point process greedy selection, and outputting the selected samples and quality level of the data set. The present application suppresses the semantic drift and redundancy of synthetic samples, and improves the semantic consistency and coverage balance of data.

The invention discloses a ship electronic information system fault prevention and control method based on backup reconstruction, and belongs to the technical field of ship electronic information system fault prevention and control. The method aims at solving the problem that dynamic balance between backup cost and reliability guarantee is difficult to achieve due to the fact that redundant resource configuration excessively depends on manual experience presetting. Comprising the following steps: establishing a target function by taking minimization of the total reinforcement cost of resource hardware as a target; the target function is solved through a rime optimization algorithm based on the dynamic inertia weight, decision variables corresponding to all resource hardware serve as rime bodies to form a rime population, all possible solutions of each rime body serve as rime particles, and an expression of the rime population based on the rime particles is obtained; a soft rime search mechanism and a hard rime puncture mechanism are combined for searching to obtain an optimal candidate solution; and an optimal solution is determined through a forward greedy selection mechanism. The method is used for electronic information system fault prevention and control.

The invention discloses a full-coverage path planning method. The method comprises the following steps: determining a grid map of an area to be subjected to path planning, and determining a transition probability matrix corresponding to the grid map; an initial population corresponding to the grid map is determined according to the transition probability matrix and preset selection strategies, and the preset selection strategies comprise a greedy selection strategy and a fuzzy selection strategy; determining a local distance contribution value of a target path in the initial population, executing a multi-modal disturbance operation on the target path according to the local distance contribution value, and generating a filial generation population corresponding to the initial population; and performing iterative optimization according to the initial population and the offspring population until an optimal path corresponding to the grid map is determined. According to the invention, technical problems of path redundancy, slow convergence speed and poor adaptability to a complex environment existing in a path planning algorithm in a complex obstacle scene in the prior art are solved.

The application discloses a kind of solid crystal machine temperature sensitive point selection method, device, equipment and storage medium, belong to equipment manufacturing technical field.The method comprises: obtaining the temperature data of first preset quantity temperature measuring points of each position point of solid crystal machine;The position point of solid crystal machine includes second preset quantity position points;Using improved AP clustering algorithm, in combination with multiple preset cluster numbers, the temperature data of each position point obtained is clustered, and the clustering result corresponding to multiple preset cluster numbers of each position point is obtained;According to multidimensional index, the clustering result corresponding to multiple preset cluster numbers of each position point is analyzed, and target cluster number is determined;Based on the clustering result corresponding to multiple preset cluster numbers of each position point, in combination with target cluster number, determine candidate point set;Temperature sensitive point set is determined from candidate point set using greedy selection algorithm.The effectiveness and accuracy of temperature sensitive point selection can be improved using the technical scheme provided by the application.

The application discloses an open source dialogue model-oriented automatic jailbreaking prompt word generation and attack method and system. The application first screens a prompt word set from an original prompt word set; secondly, based on an attack question, the prompt word with the optimal attack efficiency is screened out from the original open source prompt word set, multi-path parallel testing is carried out by using a proxy model, and the attack success rate and the prompt word length of different prompt words in the prompt word set are evaluated in real time through a dynamic evaluation mechanism based on a greedy selection strategy, the final attack prompt word is output to a target model, and after being spliced with the attack question, the attack prompt word is returned and an analysis report is output; finally, the prompt word in the analysis report is adjusted by using an automatic mutation and expert modification strategy, and the prompt word is reconstructed and iterated according to the feedback result of each round of attack, so that a new prompt word is generated. The application combines offline evolution and online decision-making to automatically generate a high-success-rate jailbreaking prompt word, controls the length and overhead, and optimizes mutation by using semantic rewriting and logic skeleton extraction.

The invention belongs to the technical field of swarm intelligence optimization algorithms, and particularly relates to a multi-strategy fusion improved star-graffiti optimization method, which comprises the following specific steps of: generating an initial population by adopting a strategy of combining Logistic chaotic mapping and mirror reflection learning, and enhancing the uniformity and quality of population distribution; in the iteration process, a nonlinear decline function and a periodic fluctuation function are used for adjusting key control parameters respectively, and dynamic balance of global exploration and local development is achieved; introducing a mode search operator based on a dynamic step size in the middle stage of iteration, and carrying out refined mining on an optimal solution region through direction search and disturbance optimization; high-quality gene transfer is accelerated through an information sharing mechanism based on a cross strategy and greedy selection. Through multi-strategy fusion improvement, the risk that the algorithm falls into local optimum is reduced, and the problem that the global search capability of an original algorithm is limited is solved.

The invention relates to the technical field of laser cutting, in particular to a parameter self-adaption method and system for laser cutting, and the method comprises the steps: constructing a solution space containing laser power and cutting speed, and initializing a plurality of nectar sources of an artificial bee colony algorithm; the follow-up error in the laser cutting process is collected in real time, the risk index of the current working condition is calculated according to the Rayleigh length, the size of the follow-up error and the change rate, and then the search step length and the fitness weight are determined; generating a new solution in the neighborhood of the nectar source by using the search step size, and calculating a fitness function value of the new solution according to the fitness weight; and the nectar sources are updated according to a greedy selection strategy, the nectar source with the highest fitness function value is selected as the optimal control parameter after iteration is completed, and the laser power and the cutting speed are adjusted according to the optimal control parameter. According to the technical scheme, the stability of a cutting energy field can be maintained, and the quality defects of slag adhering, incomplete cutting and the like are effectively overcome.

The application discloses a vehicle computing task scheduling method based on vehicle-edge-cloud cooperation and belongs to the technical field of intelligent transportation systems and Internet of Vehicles. The method is aimed at the problems of limited computing resources of intelligent networked vehicles, high task delay and large energy consumption. An improved greedy algorithm is used to optimize the task execution order, and local greedy selection is performed according to the data volume, computation volume and maximum tolerable delay of the task to output an optimal platform allocation label. Then, a feedforward neural network is trained using the labels. The trained neural network can quickly predict an optimal platform vehicle-edge-cloud cooperative allocation scheme according to the input task characteristics in a real-time scenario. Through the fusion of the algorithm and the neural network, the application realizes efficient online real-time scheduling, significantly improves the task scheduling efficiency and system reliability, has excellent real-time performance and scalability, effectively reduces the task execution delay and energy consumption, and is suitable for efficient scheduling and resource allocation of large-scale computing tasks in the Internet of Vehicles.

The application discloses an equivalent machining hole generation method for nacelle acoustic liner holes of an aero-engine and relates to the field of aero-engine manufacturing.The method comprises the following steps: a surface of a hole area of an acoustic liner is unfolded into a plane, and after the arrangement of a hole machining point of the acoustic liner is completed, the position of a multi-spindle hole machining end effector covering the hole machining point of the acoustic liner is determined to obtain an equivalent machining hole and a corresponding hole machining point set of the acoustic liner; the equivalent machining hole is generated based on a greedy selection strategy, the equivalent machining hole with a larger number of machining holes is generated preferentially, each spindle is fully utilized in a one-time positioning machining process of the multi-spindle hole machining end effector, and all equivalent machining hole sets are output; and the method can be based on acoustic liner machining of a to-be-machined hole area.

This invention proposes a deep learning-based method for joint generation of beam training and measurement patterns, belonging to the field of communication technology. The method includes: acquiring a full codebook beam measurement quality dataset; performing self-supervised pre-training on the dataset using a high-ratio random masking strategy to obtain a general beam prediction model; inputting complete beam measurement data into the model and calculating importance scores through backpropagation; iteratively selecting the beam measurement points with the highest importance scores using a greedy selection strategy based on a given beam training overhead constraint; updating the importance scores of the remaining beams after each round of selection by penalizing them based on the spatial correlation between the selected beams and the remaining beams; stopping the iteration and outputting the final optimal beam measurement pattern when the number of beam measurement points in the optimal pattern set reaches the beam training overhead constraint, guiding the base station to send the corresponding reference signal. It possesses strong generalization capabilities and can flexibly adapt to various measurement configurations and scenario changes.

The invention relates to the technical field of unmanned aerial vehicle operation of low-altitude economic intelligent logistics, in particular to a multi-distribution-center unmanned aerial vehicle distribution route planning method, which comprises the following steps of: establishing a model taking minimization of a weighted total flight distance, minimization of robust flight time and maximization of a demand satisfaction rate as decision targets; determining a distribution center by adopting a greedy selection strategy of geometric coverage and dispersity priority; and solving to obtain a Pareto frontier solution set. According to the method, the multi-target optimization model containing the flight time, the flight distance and the demand satisfaction rate is constructed, the delivery centers are selected by adopting the greedy selection strategy based on geometric coverage and dispersity, and the Pareto frontier solution set is solved by applying the NSGA-II algorithm, so that the balance of multiple decision targets is realized, the adaptability to a multi-delivery-center scene is improved, and the method has the advantages of being simple in structure, convenient to operate and high in efficiency. The distribution center selection is optimized to improve the coverage range and the service efficiency, and an optimization tradeoff scheme among the service quality, the distribution efficiency and the operation cost is provided for decision makers.

This invention provides a method for inverting and measuring the complex permittivity based on the MHHO-DE hybrid optimization algorithm, belonging to the field of electromagnetic property measurement technology. The method includes: constructing an initial Harris Hawk population of the material to be inverted within a search space, the initial Harris Hawk population comprising multiple Harris Hawk individuals, the position vectors of which represent the complex permittivity; updating the position vectors of the multiple Harris Hawk individuals based on the current optimal solution among them using a multi-strategy improved Harris Hawk optimization algorithm, forming a first optimal individual population; selecting a second optimal individual population from the first optimal individual population through a greedy selection mechanism based on mutation and crossover operations of a differential evolution algorithm; and determining the global optimal solution from the second optimal individual population, which is used to determine the complex permittivity of the material to be inverted. This method solves the technical problems of easily getting trapped in local optima and generating multivalued solutions in related technologies.

The application provides a phased array antenna beam broadening method, electronic equipment and a readable storage medium. The method comprises the following steps: constructing an optimization fitness function suitable for a phased array antenna, wherein the optimization fitness function comprises a normalized weighted combination of a beam width term for guiding an optimized half-power beam width to approximate a target beam width and a sidelobe level term for guiding an optimized sidelobe level to approximate a sidelobe suppression reference level; setting array physical parameters and beam optimization target parameters; initializing a meta-heuristic algorithm and generating an initial population, each individual in the initial population corresponding to a phase configuration scheme of all elements in the phased array antenna; starting an iterative optimization cycle, and in the case that the fitness value of each individual in the current population does not satisfy an iteration end condition, sequentially performing elite population screening, algorithm parameter updating, mutation operation, crossover operation and greedy selection operation to obtain a population of the next iteration.

The invention relates to the technical field of energy storage power station parameter optimization, and discloses a DE-based pumped storage power station water energy parameter calculation method, device and equipment, the calculation method is designed based on a differential evolution algorithm, a population is initialized firstly, each chromosome individual comprises three state variables of encoding the normal water storage level of an upper / lower reservoir and adjusting the reservoir capacity; fitness is calculated, mutation operation is carried out on the population by adopting a differential self-adaptive scaling factor strategy, and scaling factors are dynamically adjusted according to the fitness of chromosome individuals and an iteration process; performing crossover variation on the chromosome individuals according to the dynamic self-adaptive crossover probability to generate test individuals, and performing individual inspection and correction; adopting a greedy selection strategy to select individuals with excellent fitness to update the population, and judging whether the optimal individuals of the new generation population converge or not; and if convergence occurs, outputting the normal water storage level, dead water level, regulation storage capacity and power station storage capacity of the upper / lower reservoir, and returning to iteration if not convergence occurs, so as to efficiently solve the comprehensive optimal water energy parameters to provide a basis for power station design.

This invention presents a brand ambassador screening method based on efficient subset sampling. It consists of two parts: the first part constructs a graph structure G based on the social network of social media, builds a preprocessing structure for subset sampling, sets parameters, and uses greedy selection to generate a user set based on the generated reverse reachable set. Finally, the influence upper bound of the true most influential user set is calculated to provide brand ambassador candidate users based on the social network. The second part processes social network updates; when the social network is updated, the second part is used first, and then the first part is used to obtain brand ambassador candidate users. This ultimately achieves the effect of faster search for the most influential user set in the social network, allowing businesses and social media platforms to quickly identify suitable brand ambassadors.

The present invention discloses a method for collaborative path planning for a swarm of unmanned aerial vehicles (UAVs) based on the quantum Morpho mechanism. The method comprises the following steps: establishing a UAV swarm collaborative path planning model that considers variable speed and simultaneous arrival constraints; establishing a cost function for collaborative path planning for the swarm under these constraints; initializing a swarm of quantum Morpho vehicles and setting parameters; defining and calculating the odor emitted by the quantum Morpho vehicles; sorting all quantum Morpho vehicles according to the odor values ​​emitted by the quantum Morpho vehicles; sequentially executing linear and curved escape sequences, and using a simulated quantum revolving door to evolve the quantum positions of the quantum Morpho vehicles during the escape sequence. A greedy selection strategy is applied to determine the quantum position of the next generation of quantum Morpho vehicles. The evolution is terminated, and a UAV swarm path and speed matrix is ​​output. The method, while also considering variable speed and simultaneous arrival constraints under obstacle avoidance requirements, achieves rapid convergence, high convergence accuracy, and is simple to implement with a reduced number of parameters.

The application discloses a model generation method and a data malicious attack behavior detection method and device. The model generation method comprises the following steps: constructing a data malicious attack behavior detection model, the data malicious attack behavior detection model comprising a space feature extraction unit for extracting data space features, a time sequence feature extraction unit for extracting data time sequence features and an attack behavior detection unit for detecting attack behaviors, wherein the output result is an output result of the data malicious attack behavior detection model; obtaining a training data set; and training the data malicious attack behavior detection model according to training data feature information in the training data set, a data malicious attack behavior detection result corresponding to the training data feature information, a greedy selection algorithm and a neighborhood search learning algorithm of a reverse learning algorithm, to obtain a trained data malicious attack behavior detection model.

The invention relates to the technical field of aero-engines, discloses a rotor blade sorting method based on a single-parent improved genetic algorithm, and adopts the rotor blade sorting method based on the improved single-parent improved genetic algorithm. According to the method, when the target function of the initial unbalance of the impeller is constructed, the unbalance of the impeller disc and the static moment of the blades are considered, gene transposition, gene displacement and gene reverse sequence operation of a single chromosome are adopted, greedy selection and a self-adaptive variation strategy are added, iterative optimization is conducted on blade arrangement, and compared with a classic arrangement method, the method has the advantages that the method is simple and convenient to implement. And the initial unbalance amount of the impeller is effectively reduced.

The application discloses a WiFi perception-oriented minimum correlation subcarrier selection method, which comprises the following steps: collecting multi-channel CSI data and constructing a strategy candidate set; calculating a Pearson correlation coefficient matrix of candidate strategies in the strategy candidate set, constructing an average correlation matrix through the Pearson correlation coefficient matrix; based on the average correlation matrix, calculating a global correlation index and evaluating to obtain an optimal strategy index and an optimal correlation matrix corresponding to the optimal strategy index; initializing a low correlation feature index set according to the optimal correlation matrix and determining a candidate index set; performing cumulative minimization greedy selection and low correlation feature index set updating according to the candidate index set; extracting the updated low correlation feature index set to output a final low-dimensional physical feature matrix, which is a minimum correlation subcarrier.

The invention discloses a large model context generation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a to-be-queried text segment vector and a candidate text segment vector set, calculating a redundancy weight coefficient according to a budget constraint and a text segment length average value in the candidate text segment vector set, constructing a target gain function according to a similarity weight coefficient and the redundancy weight coefficient, initializing a target text segment vector set into a null set, and obtaining a target text segment vector set. And sequentially selecting the text segment vectors which enable the target gain function to be maximum from the candidate text segment vector set by utilizing a greedy algorithm, adding the text segment vectors into a target text segment vector set to obtain a target text segment vector set after greedy selection, and generating a context corresponding to the to-be-queried text segment vector according to the target text segment vector set after greedy selection. By adopting the scheme of the invention, the redundancy between the selected text segments can be reduced, and the diversity and the utilization rate of evidences can be improved, so that the quality and the consistency of the generated context can be ensured.

The invention discloses a fair incentive method based on dynamic federated learning, and belongs to the technical field of federated learning. Aiming at the problem of how to reduce the negative influence of heterogeneous client data on the performance of a federal global model while ensuring the fairness of a federal learning environment, a training contract algorithm based on a multi-dimensional bidding mechanism is designed, so that an illegal client is prevented from generating a contract for the federal global model. A similarity smoothness measurement mechanism based on model parameters is put forward to fairly measure contribution of clients, a concept of training value coefficients is put forward, an exclusive training value coefficient is created for each client by means of a reputation value mapping mechanism, and the value and necessity of the clients in subsequent learning are measured. And distribution and efficient utilization of resources in a subsequent learning process are promoted. A client self-adaptive adjustment algorithm based on greedy selection is provided, the algorithm fuses a greedy thought and a feedback control mechanism, a fair and efficient motivation target is achieved, and sustainability and robustness of federal learning ecology are guaranteed.

The invention discloses a pathological image classification method and system based on class incremental learning in the technical field of medical images.The method comprises the steps that a greedy selection algorithm based on the maximum mean value difference is adopted, a playback sample set is constructed from samples in the current stage, and when a next incremental task is entered, the playback sample set is replayed; new data of the stage and a playback sample stored in the previous stage are input into the model together for joint training, in the training process, class balance loss of prior regularization is introduced, meanwhile, knowledge distillation is conducted on model output of the current stage and model output of the previous stage, a gradient calibration mechanism is introduced into a full-connection classification layer, and therefore the model output of the current stage and the model output of the previous stage are obtained. Dynamically adjusting the gradient change of the new and old categories; the method can effectively solve the problem that the performance of the pathological image is reduced due to domain difference and class imbalance in the class incremental learning process, maintains the classification accuracy of the existing disease classes, has the capability of quickly adapting to new disease classes, and has important clinical diagnosis application value.

The present application relates to a kind of solving multi-class agricultural machinery sequential task optimization scheduling method.The present application is for the preparation time of sequential task agricultural machinery scheduling problem, with the maximum operation time of agricultural machinery as scheduling goal.Minimize.The algorithm of the present application is based on the operation order coding of agricultural machinery to farmland, uses improved MNEH to generate initialization population, and is arranged in ascending order according to fitness value, wherein the first third is as the leader of each population, and the last two-thirds is as the follower of each population.The leader of each sub-population searches neighborhood individuals using serial neighborhood search strategy;the follower uses parallel neighborhood search strategy to generate neighborhood individuals, and uses greedy selection mode to select, if the follower is better than the leader, the two are exchanged;local search algorithm is designed in the algorithm to act on optimal individual, and age variable is designed to record the leader update condition of each population, if the age of a certain leader reaches certain limit, new individual replaces the individual and is placed into population.