133 results about "Evolutionary computation" patented technology

The invention involves generating and presenting, typically electronically, a number of design alternatives to persons who are participating in the design, selection, or market research exercise. The participants (referred to as “selectors”) transmit data indicative of their preferences among or between the presented design alternatives, and that data is used to derive a new generation of design alternatives or proposals. The new designs are generated through the use of a computer program exploiting a genetic or evolutionary computational technique. The process is repeated, typically for many iterations or cycles.

A system, methods and apparatus are described involving the self-organizing dynamics of networks of distributed computers. The system uses intelligent mobile software agents in a multi-agent system to perform numerous functions, including search, analysis, collaboration, negotiation, decision making and structural transformation. Data are continuously input, analyzed, organized, reorganized, used and output for specific commercial and industrial applications. The system uses combinations of AI techniques, including evolutionary computation, genetic programming and evolving artificial neural networks; consequently, the system learns, anticipates and adapts. The numerous categories of applications of the system include optimizing network dynamics, collective robotics systems, automated commercial systems and molecular modeling systems. Given the application of complexity theory and modal and temporal logics to self-organizing dynamic networks, a novel model of intelligent systems is presented.

A system, method and program product for selecting an algorithm and feature set to solve a problem. A perpetual analytics system is disclosed that provides a genetic algorithm for jointly selecting an algorithm and feature set to solve a problem, comprising: an evolutionary computing engine for processing data encoded as chromosomes, wherein each chromosome encodes an algorithm and a feature set; a domain knowledge store that maintains a plurality of algorithms and a plurality of features; a system for applying a generation of chromosomes to a set of data to provide a set of results; and a fitness function for evaluating the set of results to rate a performance of each chromosome in the set of chromosomes; wherein the evolutionary computing engine is adapted to evolve a subset of the set of chromosomes into a new generation of chromosomes.

Cardiac Output (CO) has traditionally been difficult, dangerous, and expensive to obtain. Surrogate measures such as pulse rate and blood pressure have therefore been used to permit an estimate of CO. MEMS technology, evolutionary computation, and time-frequency signal analysis techniques provide a technology to non-invasively estimate CO, based on precordial (chest wall) motions. The technology detects a ventricular contraction time point, and stroke volume, from chest wall motion measurements. As CO is the product of heart rate and stroke volume, these algorithms permit continuous, beat to beat CO assessment. Nontraditional Wavelet analysis can be used to extract features from chest acceleration. A learning tool is preferable to define the packets which best correlate to contraction time and stroke volume.

The invention discloses a multi-target reactive power optimization method for an electric system, which belongs to the field of reactive power optimization for electric systems. The method includes: modifying the Memetic algorithm to adapt to multi-target optimization, applying the modified Memetic algorithm to the problem of multi-target reactive power optimization for the electric system, and working out a Pareto optimal solution of the multi-target problem; and judging whether algorithm convergence conditions are met or not, and if yes, completing optimization and outputting optimization results. The multi-target reactive power optimization method has the advantages that the algorithm for solving the problem of multi-target reactive power optimization is provided, the method is more suitable for solving the multi-target problem while giving play to existing advantages of the Memetic algorithm which integrates local searching and evolutionary computation and has high global search capacity and the like, and searching efficiency is improved while algorithm robustness is improved.

The present invention provides an order sequencing optimization method for logistics. The process of the method includes the following steps that: since the same type of commodities may be stored in a plurality of different containers, containers from which corresponding commodities in orders should be taken out are determined according to the expiration dates of the commodities in the orders, the number of remaining commodities in the containers where the commodities are located or the distances of the containers before order sequencing optimization is performed; and the orders of the commodities of which the containers are determined are optimally sequenced. According to the order sequencing optimization method for logistics of the invention, an evolutionary computation optimization method is adopted to optimally sequence the orders, so that the orders of the same commodities can be arranged at adjacent positions as much as possible, and therefore, commodities which have taken out can be repeatedly utilized, and the problem of low efficiency caused by frequent transport of the commodities can be avoided; and a constructed learning group concept is utilized to expand the search range of the algorithm, so that optimization calculation can be prevented from being trapped in local optimum, and convergence precision is greatly increased. With the order sequencing optimization method for logistics of the invention adopted, the optimal commodity picking efficiency of the commodities in the orders in a logistics process can be realized.

A hybrid router for dynamical control systems is described. The mobile hybrid software router (MHSR) combines distinctive computational and mathematical techniques, including evolutionary computation (EC), probabilistic simulations (PS), machine learning and artificial neural networks (A-NNs), in order to solve unique problems encountered in an unknown environment in real time. Embodied in intelligent mobile software agents (IMSAs), the MHSR operates within a multi-agent system (MAS) to continually optimize system operation. The MHSR is applied to several major complex system categories. In one embodiment of the system, the MHSR is implemented in hardware, including continuously programmable field programmable gate arrays (CP-FPGAs), for perpetually reconfigurable evolvable hardware operation. Whether in application-specific or multi-functional mode, the MHSR is useful to groups of agents in intelligent systems for adaptation to uncertain environments in order to perform self-organization capabilities.

The invention relates to a new face-iris combination identifying method-characteristic layer extraction and combination. A face-iris characteristic extraction layer combining system is established according to nerve network, evolution calculation and fuzzy theory. For structure design, full and local geometry topological structure is adopted. A particle-group optimizing arithmetic is utilized to optimize network control parameters. When the characteristics of the face and the iris image are extracted, techniques of a super-resolution image reinforcing arithmetic, an illumination compensating arithmetic based on improved spherical harmonic function, gesture estimation based on linear relevant filters, Candide model based on a three-dimensional face and expression analysis based on an ASM arithmetic, etc., are adopted to robustly extract the eigenvectors of the face and the iris, and a self-developed double face-iris collecting device is also adopted to collect images of the face and the iris image. The method not only can establish a new system which is provided with learning capability and can automatically choose optimal network topological structure and automatically regulate net control parameters, but also can overcome and reduce the bad impacts of factors of environment and physiology, etc., during the extraction process to the independent characteristics of the face and the iris, thus effectively enhancing the identifying rate of the face-iris combination identification and promoting the system performance based on the face-iris combination identification to develop towards practical, reliable and acceptable directions.

In an output controlling method for controlling output of a driving power source or motor installed in a vehicle, the relationship between a primary output-controller, which is manipulated by the user, and a secondary output-controller, which directly operates the motor, is regulated by control parameter subjected to evolution by using evolutionary computing based on at least one of the following: the user's characteristics, driving conditions, environmental changes, and deterioration of the drive power source with time. The evolution is conducted on-line or on a real-time basis. The primary output-controller includes an acceleration pedal or grip, and the secondary output-controller includes a throttle valve for an internal combustion engine or a running current controller for an electric motor.

The present invention relates to a method for controlling motions of a robot using evolutionary computation, the method including constructing a database by collecting patterns of human motion, evolving the database using a genetic operator that is based upon PCA and dynamics-based optimization, and creating motion of a robot in real time using the evolved database. According to the present invention, with the evolved database, a robot may learn human motions and control optimized motions in real time.

The invention relates to a grid resource allocation method based on a parallel genetic algorithm. The method comprises the following steps: firstly, the information is initialized in a main thread, such as task collection, machine collection, an execution time matrix E of the task, and mapping of a sub-task to the machine, etc.; then a plurality of sub threads are generated and mapped to different processors, an initializing sub-population is independently generated by each sub thread, evolutionary computation is performed in parallel, the optimum individual of each generation is transferred to the main thread, the main thread performs comparison, and the optimum individual is retained; when the predetermined generation arrives, the transfer operation between the sub-groups is performed; and the operation of the main thread and all the sub-groups cannot be finished until the termination conditions are met. The genetic algorithm is taken as the most effective heuristic global stochastic searching method, and the solution of the NP problem can be performed. The quality and the speed for the algorithm for solving are improved by the parallel genetic algorithm proposed according to the natural parallelism of the genetic algorithm, and the method is an effective grid energy resource optimization method and favorable for improving the service quality of the grid.

An artificial multiped is constructed (either in simulation or embodied) in such a way that its natural body dynamics allow the lower part of each leg to swing naturally under the influence of gravity. The upper part of each leg is actively actuated in the sagittal plane. The necessary input to drive the above-mentioned actuators is derived from a neural network controller. The latter is arranged as two bi-directionally coupled chains of neural oscillators, the number of which equals twice that of the legs to be actuated. Parameter optimisation of the controllers is achieved by evolutionary computation in the form of a genetic algorithm.

This invention relates to a non-linear correction method of linear frequency-modulated radar object meter, which belongs to signal process and automatic measurement technique field. It divides the frequency-scanning voltage curve of the control oscillator and uses frequency spectrum analysis to extract characteristics information and adopts principle of evolution computation to get the end point coordinates of the curve to form non-linear frequency-scanning curve to compensate the non-linear property of the control oscillator without measurement of high-frequency microwave signals or phase to generate the linear frequency-modulated signals with high linear property. It solves the problem of linear frequency-modulated radar non-linear property.

The invention discloses a method for expansion planning of a power transmission network, which relates to the technical field of power transmission network plan and evolutionary computation. The method comprises the following steps of: initializing parameters and acquiring an initial position and an objective function value of each monkey in a monkey group; executing the crawling process of the monkey group according to the initialized parameters, the initial positions and the objective function values; executing the climbing process of the monkey group; judging whether a termination criterion is satisfied, if so, ending the process; otherwise, executing the crawling process. Because the method adopts two crawling processes with strides and small steps and the climbing process, and introduces a gazing jump process, a cooperative process and a random disturbance mechanism, the method overcomes the defect that a monkey group algorithm can only solve the problem of continuous optimization, and solves mixed-integer optimization problems of the expansion planning of the power transmission network; besides, the method has the advantages of simple implementation, quick computation speed and stronger robustness.

The present invention provides a method for creating a process change detection algorithm. An evolutionary computing technique is applied to at least one process dataset containing at least one known process change. The evolutionary computing technique will generate a process state function (or a scaling coefficient set for use with an existing process state function) that optimizes detection of the known process changes. The generated process state function or coefficients can then be applied thereafter to future datasets (either in real-time or after processing) to detect process changes.

The invention discloses a neural network search distributed training system and training method based on evolutionary computation, and solves the problem of time consumption of an evolutionary neuralnetwork search algorithm. A neural network search distributed training system based on evolutionary computation comprises a server node and at least one computing node, and the server node communicates with the computing nodes; the server node is used for generating each generation of population (one individual represents one neural network model) based on an evolutionary algorithm, asynchronouslyreceiving a request sent by the computing node, distributing to-be-evaluated individuals, receiving the individuals of which the computing node completes fitness calculation, and controlling the computing node at the same time; and the computing node is used for asynchronously sending a request instruction to the server node, receiving a data packet (including an individual to be evaluated), decoding the individual to generate a neural network, training and verifying on a specified data set, completing neural network performance evaluation, namely completing individual fitness evaluation, andfeeding back the evaluated individual to the server node.

The invention provides a method and apparatus for predicting dissolved oxygen in aquaculture. The method comprises: collecting dissolved oxygen data of an aquaculture pond; carrying out ensemble empirical mode decomposition on the dissolved oxygen data to obtain a high-frequency term, an intermediate-frequency term, a low-frequency term and a residual component; carrying out modeling on the high-frequency term and the intermediate-frequency term by using a least square support vector machine and carrying out dissolved oxygen prediction; carrying out modeling on the low-frequency term by using a mind-evolutionary-computation-based optimized BP neural network and carrying out dissolved oxygen prediction; carrying out modeling on the residual component by using a gray forecasting model and carrying out dissolved oxygen prediction; and superposing prediction results of the least square support vector machine, the mind-evolutionary-computation-based optimized BP neural network, and the gray forecasting model to obtain a prediction value of dissolved oxygen in aquaculture. Therefore, the changing situation of the dissolved oxygen can be obtained accurately; and because the high-frequency term, the intermediate-frequency term, the low-frequency term and the residual component are predicted by using different dissolved oxygen prediction methods, accurate prediction of dissolved oxygen data in aquaculture can be improved.

The invention discloses an evolutionary computation-based method for diagnosing compressor fault. The method comprises the following steps: data about the working condition of a compressor is collected and acquired; an observation point is taken as a node, and the correlation between observation points produced by the actual compressor physics; a weighting undirected graph G(V,A) is set up for measured data; the distance between fault data points is gauged by means of a relational graph instead of the traditional Euclidean distance method; the data about the fault of the compressor is converted into a problem with the partitioning of the weighting undirected graph G; the Pareto Principle structural map is adopted for partitioning cost function; an evolutionary calculation chart partitioning computation based on the improvement on the complexity system theory is designed to partitioning the relational graph; and the clustering / classification of the compressor fault data is realized through the optimal computation of the cost function, so that the purpose of fault diagnosis can be achieved.

The invention belongs to the two major fields of ship motion parameter identification and evolutionary computation, in particular to a ship lateral three-degree-of-freedom motion parameter identification method based on a multi-objective evolutionary algorithm. The method comprises the following steps of: establishing a state equation and an observation equation model according to a differential equation of lateral three-degree-of-freedom motion of the ship; according to the experimental data, establishing the model of disturbance force and disturbance moment,with taking Manhattan distance asthe criterion, evaluating the error between the observed value and the output value of the model, and establishing the multi-objective output model of ship lateral motion, optimizing the multi-objective optimization function of ship lateral motion by evolutionary algorithm, and giving the optimal pareto front, selecting the parameter corresponding to a point from the front surface as the output optimal parameter. There is no need to consider the weighting factors between the motions of degrees of freedom. A series of non-dominant solutions can be obtained, and the most suitable set of transverse hydrodynamic parameters can be selected from these solutions according to customer preferences. The experimental data are few, the efficiency is high and the error is small.

The invention discloses a multicast routing optimization method based on quantum evolution, relating to the evolutionary computation field and aiming to solve the problems of long optimization time, easy convergence to local optimum, and the like existing in the prior art. The multicast routing optimization method based on quantum evolution comprises the following steps: 1. generating a stochastic network and setting operational parameters; 2. solving all alternative paths satisfying time-delay conditions for each destination node; 3. carrying out quantum coding on the alternative path set of each destination node so as to acquire a state matrix; 4. carrying out quantum observation on the state matrix so as to obtain a set of binary strings; 5. decoding the binary strings so as to obtain the selected path of each destination node, and computing a multicast tree fitness function; 6. carrying out quantum variation on the state matrix by a quantum revolution door; 7. obtaining a set of new binary strings by observing the state matrix, and then computing a new multicast tree fitness function after the new binary strings are decoded. The invention has low complexity of computation and low cost of the optimized multicast tree and can be used for effectively allocating network resources.

The invention discloses an RFID network topology method based on the multi-agent evolutionary algorithm and belongs to the field of automatic control and information technology. The method combines a multi-agent system with evolutionary computation to solve the RFID network topology problem. The method is characterized in that each agent in an agent grid is initialized according to two algorithms, and then a neighborhood competition operator, a neighborhood right-angled crossover operator, a mutation operator and a self-learning operation are designed to optimize the agents. Verification results show that the method has great advantages in terms of solving the maximum coverage rate of reader-writers, the minimum number of the reader-writers and the minimum interference rate of the reader-writers when used for evaluating and solving the RFID network topology problem and is an effective method for solving the RFID network topology problem.

The invention discloses a method for automatically extracting a traditional Chinese medicine pulse characteristic function. The method comprises the following steps: obtaining periodic data related to time series from source beyond a computer; selecting data from the periodic data, normalizing and automatically building a mathematical model for the selected data; extracting a characteristic function which meets requirements and characterizes the data characteristic. The method is different from the common methods such as multiple regression method, wavelet analysis method or neural network method and the like. The characteristic function extracted by the method has an unlimited concrete form, an characteristic function extraction program is based on an evolutionary computation method, and the evolutionary computation is a self-organizing and self-adapting artificial intelligence technology which simulates the biological evolution process and a problem-solving mechanism, and is a searching algorithm having a 'generation-detection' iterative process, and the extracted characteristic function can assist doctors of traditional Chinese medicine in objectivizing and quantizing the diagnosis work.

The invention discloses a federated learning method and device based on evolutionary computation, a central server and a readable storage medium. The method comprises the following steps of: acquiringa hyper-parameter of each participant group, and sending the hyper-parameters to each participant of each participant group to enable the participants in each participant group to carry out federatedlearning; acquiring sub-target models obtained by trained through federated learning of the participant groups, and performing iterative evolutionary computation based on the performance index of each sub-target model and the hyper-parameter corresponding to each participant group to obtain a target hyper-parameter; and sending the target hyper-parameter to each participant participating in federated learning to enable the participants to determine respective initial model parameters based on the target hyper-parameter and perform federated learning based on the initial model parameters to obtain a target model. The hyper-parameter optimization efficiency in federated learning is improved through evolutionary computation, and meanwhile, the performance of a federated learning model is remarkably improved.

The invention discloses a federated learning method and devicebased on evolutionary computation, equipment and a readable storage medium. The method comprises the following steps of: determining hyper-parameters of a first group corresponding to the first equipment in a hyper-parameter set, whereinthe equipment in the first group performs federated learning based on the hyper-parameters of the first group; obtaining a first sub-target model corresponding to the first group, receiving performance indexes of a second sub-target model sent by a plurality of pieces of second equipment, performingiterative evolution calculation based on the performance indexes of the first sub-target model, the performance indexes of the plurality of second sub-target models and the hyper-parameter set, wherein the first equipment and the plurality of second equipment determine respective initial model parameters based on the target hyper-parameters, and the first equipment and the plurality of second equipment perform federated learning based on the respective initial model parameters to obtain a target model. The performance of the federated learning model is remarkably improved while the hyper-parameter optimization efficiency in federated learning is improved.

The invention discloses a resource-constrained project scheduling method based on a multi-agent evolutionary algorithm, belonging to the technical field of automatic control and information. According to the method, a multi-agent system is combined with evolutionary computation so as to solve the resource-constrained project scheduling problem. The method is characterized by comprising the following steps of: initializing each agent in an agent grid according to two algorithms; and designing a neighborhood competition operator, a neighborhood cross operator, a mutation operator and a self-learning operator for optimizing the agent. The verification result proves that the method has advantages in two aspects for evaluating the effectiveness of the method for solving the resource-constrained project scheduling problem, namely the solving of the optimal solution proportion and the solving of the average error deviating from the optimal solution, so that the method effectively solves the resource-constrained project scheduling problem, and the method can be expanded to solving other combined optimization problems with precedence relationship constraints.

Embodiments of the invention provide an adversarial sample generation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining an original text; determining a replacement word candidate set of each word in the original text; and based on a particle swarm optimization algorithm, searching a sample of the attack target model from a discrete space formed by the combination of the replacement word candidate sets, and generating an adversarial sample. According to the embodiment of the invention, the particle swarm optimization algorithm is used forsearching the adversarial sample, and the particle swarm optimization is more efficient than the genetic algorithm as a meta-heuristic group evolution calculation method, so that the search speed canbe increased when the algorithm is used for searching the adversarial sample, and the attack success rate can also be increased. For different natural language processing models, the embodiment of theinvention can quickly and efficiently generate a large number of high-quality confrontation samples, successfully cheat the target model and further expose the vulnerability of the target model, andhas good practicability.

The invention discloses an adaptive echo cancellation method. The method introduces the viewpoint of evolutionary computation in the artificial intelligence, simultaneously employs the cross alternative method and the idea of the function global optimization and combines the idea of time driving, so as to effectively solve the problem that the constant Delta value of the parameter of the conventional NLMS algorithm can not be determined, thus improving the performance of the echo cancellation algorithm NLMS.