31 results about "Evolutionary learning" patented technology

A Self-Designing Intelligent Signal Processing System Capable of Evolutional Learning for Classification/Recognition of One and Multidimensional Signals is described which classifies data by an evolutionary learning environment that develops the features and algorithms that are best suited for the recognition problem under consideration. The System adaptively learns what data need to be extracted in order to recognize the given pattern with the least amount of processing. The System decides what features need to be selected for classification and/or recognition to fit a certain structure that leads to the least amount of processing according to the nature of the given data. The System disclosed herein is capable of recognizing an enormously large number of patterns with a high accuracy.

A methodology is presented to address the need for rapid generation and optimization of algorithms that are efficient in solving a given class of problems within the framework of a software environment. The environment incorporates an evolutionary learning methodology which automatically optimizes the configurations of procedural components of the algorithm. In this way, both the efficiency and the quality of algorithm development is enhanced significantly.

The invention relates to a method for automatically learning and extended evolution about the description rules of Chinese basic blocks, and its characteristics are as follow: it rapidly excludes the unreliable description rules of the different stages to greatly increase the processing efficiency for automatically accessing rules through introducing a confidence evaluation mechanism. It opens a learning environment and breaks the constraint of automatically enhance the learning capacity for lack of training corpus through the introduction of external knowledge of the rules of evolutionary learning process for different languages. It continually introduces additional internal and external linkages vocabulary context restrictions knowledge from the original series described marker rules to dynamic-devise the state space of positive and negative cases, so that it gradually evolves the structure rules to form a hierarchical, multi-block sized of basic rules description system.

The invention provides an evolutionary learning method for intelligent monitoring of cutter states. A three-way acceleration sensor and a microphone are utilized to collect vibration signals and acoustic signals, the signals carries out smoothing, and the signals are divided into a training set and a test set; deep-level features of the dynamic signals are automatically extracted using a stacked auto-encoder, and extracted features are classified; according to the accuracy of a training set model, an algorithm carries out weight distribution, the final predicted cutter state is obtained through weighted averaging, and model related parameters are saved; and the real-time vibration signals and the acoustic signals in the actual machining process are preprocessed and then input into a savedmonitoring model, the cutter state of the corresponding signals is obtained, the data label with the higher confidence level is stored, and the network parameters are updated so as to realize the evolution learning of the cutter state intelligent monitoring. According to the method, manual participation can be avoided, the calculation complexity is reduced, and the influence of machine cutter performance degradation on the prediction accuracy of the cutter state monitoring model can be weakened.

The invention discloses a channel network steady-state evolutionary game method in a blockchain environment. Based on finite rationality and a replication dynamic mechanism of a biological evolution process, node transactions in the channel network are modeled into a dynamic evolutionary game model, the model designs a channel network dynamic defense mechanism based on evolutionary game, the mechanism adds a defense strategy to a strategy space of an I-type node, and under the condition of considering the attack cost and success rate of the II-type node under the condition of deploying the defense strategy, and the defense mechanism can help the I-type node to resist attack behaviors of the II-type node at a certain probability. The nodes of the model have limited and rational evolutionarylearning capability, and can dynamically adjust respective strategies to achieve effective defense according to different attack strategies of attackers.

The invention discloses a virus detection method based on a collaborative immune network evolutionary algorithm, and belongs to the technical field of network security. According to the method, detectors in the immune network are optimized continually through the mutual collaboration among various immune cells. The method introducing a non-self set in the evolutionary process, and performing clonal selection on mature detectors based on the detector fitness to the non-self set; simultaneously, updating mutation methods with mutation step size self-adaptation and capable of changing mature detectors through an evolutionary algebra through the evolutionary algebra, and raising a network inhibition strategy based on concentration partition, thus, the network cell diversity is improved, and the redundancy rate of detectors is reduced simultaneously. According to the virus detection method based on the collaborative immune network evolutionary algorithm, advantages of the evolutionary algorithm and the artificial immune technology are combined and fully used, and the network virus detection efficiency is improved.

The invention discloses a multi-virtual-power-plant dynamic game transaction behavior analysis method based on finite rationality, and the method comprises: firstly enabling each virtual power plant bidding main body to fully consider the target demands of the virtual power plant bidding main body at the development stage, and researching the dynamic pricing behaviors of an upper-layer operator; secondly, performing transaction behavior analysis on the multi-virtual power plant by adopting different transaction target modeling, performing evolutionary learning on finite rational transaction behavior information by adopting a particle swarm algorithm, and further improving a self target by learning a competitor strategy so as to gradually optimize the self target; and finally, researching amulti-virtual power plant dynamic game calculation process, and proposing a dynamic game particle swarm optimization algorithm to be combined with an optimization toolbox to solve the game model. Thesolution of the proposed dynamic game model has good convergence, and provides new ideas and references for the virtual power plant to formulate different transaction targets to participate in markettransactions.

The invention discloses an unmanned aerial vehicle dunking and rebounding self-evolution intelligent system and control method. The unmanned aerial vehicle intelligent system comprises a four-rotor unmanned aerial vehicle, three cameras, a flight controller, an airborne central processor, a ground data processing base station and a basketball shooting device. The control method comprises a controlmethod of a double-under-stability system and a control method for self-evolutionary learning of the unmanned aerial vehicle. According to the control method of the double-under-stability system, estimated interference is added to an unmanned aerial vehicle under-stability flight model, an interference compensator is designed for approximation in order to compensate the counter-acting force of the estimated interference on the unmanned aerial vehicle, so that unstable factors caused by the estimated interference to the unmanned aerial vehicle can be eliminated, the closed-loop performance ofthe whole double-under-stability system can be enhanced, and the self-repairing of the flying state of the unmanned aerial vehicle can be realized. The control method for self-evolutionary learning ofthe unmanned aerial vehicle is a control method obtained by combining a mechanism model and a data model, wherein the mechanism model is used for establishing rules for a flight track of a basketball, and the data model is used for carrying out learning training through a certain number of samples.

The invention provides a molecular structure acquisition method and device, electronic equipment and a storage medium, and relates to the field of artificial intelligence such as deep learning, and the method comprises the steps: for an initial seed, executing the following first processing: generating M molecular structures according to the seed, M being a positive integer greater than one; taking the M molecular structures as to-be-selected molecular structures, and selecting part of molecular structures from the to-be-selected molecular structures as offspring molecular structures; and performing evolutionary learning on the offspring molecular structure, taking the offspring molecular structure after evolutionary learning as a seed, repeatedly executing the first processing until convergence to an optimization target, and when convergence to the optimization target, taking the newly selected molecular structure as the required molecular structure. By applying the scheme provided by the invention, the implementation cost can be reduced, and the processing efficiency can be improved.

A method of evaluating business processes comprises inputting a set of initial processes, inputting a distance function, and determining whether new processes are allowed. If such new processes are not allowed, the method determines which of the initial processes is the best process by applying the initial processes to the distance function to determine which of the initial processes has the lowest measure score produced by the distance function. Therefore, the method identifies the initial process having the lowest measure score as the best-of-breed process. If such new processes are allowed, the method determines which of the initial processes and the new processes is the best using the following process. The process of finding the best process translates the initial processes to counterparts for use with an evolutionary algorithm and selects a fitness function for the evolutionary algorithm. This process continues by applying the evolutionary algorithm to the counterparts using the fitness function to generate an output state (score) and determining which of the processes is closest to the output state to identify the best process. Then the best-of-breed process can be translated and output to the user.

A method of evaluating business processes comprises inputting a set of initial processes, inputting a distance function, and determining whether new processes are allowed. If such new processes are not allowed, the method determines which of the initial processes is the best process by applying the initial processes to the distance function to determine which of the initial processes has the lowest measure score produced by the distance function. Therefore, the method identifies the initial process having the lowest measure score as the best-of-breed process. If such new processes are allowed, the method determines which of the initial processes and the new processes is the best using the following process. The process of finding the best process translates the initial processes to counterparts for use with an evolutionary algorithm and selects a fitness function for the evolutionary algorithm. This process continues by applying the evolutionary algorithm to the counterparts using the fitness function to generate an output state (score) and determining which of the processes is closest to the output state to identify the best process. Then the best-of-breed process can be translated and output to the user.

The invention discloses a digital image filter circuit design method based on FPGA evolutionary learning. Characteristic optimization is carried out a circuit code and a (2+lambda) evolutionary strategy (ES) by means of a gene expression, a proper relation operation set is evolved through a multi-target model through effect learning before filtering of sample images and after filtering of the sample images, an image filtering effect can be obtained through a filter circuit designed in the mode as good as possible, in a learning phase, a filter can obtain an optimized logic composition structure after definitive evolutionary algebra, and a hardware circuit with image filtered is achieved on an FPGA chip through VHDL conversion and competition and hazard elimination design. According to the method, the non-linear filter circuit is obtained through evolution, so that filtered images are clear and edges of the images are clear.

The invention discloses an intelligent robot and a learning method thereof, and the intelligent robot comprises an intelligent agent, and the intelligent agent comprises a brain model which is used for obtaining a first interaction message, and a state fed back by an external environment and the intelligent agent, and outputting an action and/or a second interaction message according to the state and/or the first interaction message, obtaining a new first interaction message, and/or enabling the external environment to adjust the state according to the action; the heart model is used for outputting a return according to at least one of the state and the first interaction message; the updating module is used for updating the brain model according to the return so as to realize intelligent agent learning; and the heart model is also used for carrying out evolutionary learning and updating the brain model by utilizing the evolutionary heart model so as to obtain a heart model and a meninx type which are suitable for the environment. According to the intelligent robot, the brain model is driven to learn based on the heart model of the intelligent agent, and the heart model of the intelligent agent is updated in group evolution, so that the intelligent robot can solve problems in complex fields.

The invention discloses a distributed resource dynamic allocation method based on an evolutionary game theory, which comprises the following steps: S1, judging whether supply and demand of a physicalmachine and a user task are matched, if yes, executing the step S2, and if not, executing the step S3; S2, performing a game between the physical machine and the user task based on the price bargaining strategy of the fraudulent sound potential behavior, judging whether the transaction between the physical machine and the user task can be successful, and if yes, executing the step S4; otherwise, executing the step S3; S3, acquiring a new physical machine or user task, and executing the step S1; and S4, determining a resource exchange price ratio, and allocating resources to the user task. According to the method, a bargaining phenomenon commonly existing in real life is added in a game link; according to the method, the fitness function is combined, so that the main body repeatedly plays agame with the purposes of resource balance and efficient resource utilization in the game link, the strategy of individuals with higher fitness in similar main bodies is learned in the evolutionary learning link, and the excellent performance of the group in the aspects of configuration efficiency, resource balance and the like is continuously promoted.

The invention discloses a multi-machine cooperative control method and system based on negentropy increase, and the method comprises the steps: employing a transfer learning theory, researching the dynamic evolution rule of multi-dimensional quality state spatial distribution, carrying out the dynamic prediction of a multi-dimensional quality state fluctuation rule in an intelligent manufacturing process, searching for influence factors capable of enabling the entropy value of the community network of intelligent manufacturing to be smaller, and then revealing the rule capable of enabling the fluctuation of the quality state to be smaller, so as to form a negative entropy increasing method for the multi-dimensional quality state in the intelligent manufacturing process; and then introducing a micro-evolution learning mechanism, constructing a multi-machine cooperative self-adaptive control method, completing structure adjustment control through a negentropy controller, and completing process adjustment control through self-adaptive control such that the dynamic cyclic self-adaptive control over the multi-dimensional quality state of the intelligent manufacturing process is achieved. The method and the system can be widely applied to the field of product quality guarantee of intelligent manufacturing.

The invention provides an information prediction method and system based on an autonomous evolution learner, and the method and system can find the required data dimension and effective model networkstructure for a model based on an autonomous training method, automatically complete the training of each model, and accelerate the optimization process of various parameters of the model through thecombination of autonomous training and heredity. According to the method, the problem of data dimension disasters in information mining is solved in a multi-model training mode.