36results about How to "Improve search performance" patented technology

Provided is a multi-target route planning device and method considering searching robot target node timeliness. A multi-target genetic algorithm is used to achieve route planning considering two optimization objectives including route consumption and the node timeliness through two stages of decoupling which comprise the route planning and route generating, and therefore improvement of searching performance of a mobile robot which is supervised by an operator is benefited. Particularly, when the problem of the node timeliness needs to be considered in the process of route searching, a good searching manifestation can be obtained.

The invention discloses a keyword extracting method based on Word2Vec and a Query log, and relates to the field of information processing. The method includes the steps of S1, building a specific word list of a target field; S2, obtaining candidate keywords of documents in a document set; S3, obtaining word vectors of a plurality of dimensions of each candidate keyword; S4, calculating the cosine similarities between the word vectors of any candidate keyword L and a center vector, judging whether the candidate keyword L exists in the specific word list or not, if the candidate keyword L exists in the specific word list, directly implementing the step S5, and if the candidate keyword L does not exist in the specific word list, directly implementing the step S6; S5, multiplying the obtained cosine similarities by a weighting factor i to obtain new cosine similarities, and implementing the step S6; S6, ranking the values of the cosine similarities from large to small, outputting the values of m cosine similarities from the cosine similarity with the largest value, and obtaining final keywords. By means of the keyword extracting method, the keywords with the ideal quality can be rapidly and efficiently extracted for texts in specific fields, oral words are prevented from being introduced, and the extracted keywords are high in quality.

The invention discloses LSTM hub single product energy consumption prediction based on incremental clustering, and relates to the technical field of hub single product energy consumption prediction. According to the method, clustering analysis of the hub characteristic parameters is achieved through dynamic incremental density clustering based on PCA, the type of a historical product to which a newly-added product belongs is obtained, and a Pearson coefficient and Adaptive-Lasso algorithm are utilized on the basis of an energy consumption influence factor system for screening out single product energy consumption high interpretability factors; by means of BP, prediction of new product strong interpretability variables is achieved, an LSTM energy consumption prediction model is constructedfor each cluster product, effective prediction of new product unit consumption is achieved, optimization of LSTM is achieved through ADE, and meanwhile an incremental learning strategy is introduced to achieve dynamic updating of the model. The effectiveness of the prediction method is verified, the root-mean-square error RMSE of energy consumption prediction is lowered to 0.016524, compared withother algorithms without incremental learning, the root-mean-square error RMSE is lowered by 0.013653 on average, meanwhile, the ADE search performance of the algorithm is good, and the RMSE of a training set is reduced by 0.004089 compared with that of the average of DE-LSTM with incremental learning , and the operation time is effectively shortened.

The invention provides a determination method and a determination device for semantic redundancy and a corresponding search method and device. The determination method for semantic redundancy comprises the steps of S1, determining a word A and a collocation word B thereof; S2, making the statistics of a context vector of a collocation word pair composed of the word A and the word B from corpus and a context vector of the word A; S3, computing the similarity between the context vector of the collocation word pair composed of the word A and the word B and the context vector of the word A, if the similarity is larger than a preset similarity threshold, determining that the collocation word pair composed of the word A and the word B and the word A form a semantic redundancy pair, wherein the word B is a redundancy word. By using the determination method and the determination device for semantic redundancy and the corresponding search method and device, the semantic redundancy condition in query is effectively determined, a basis is provided for the elimination of redundancy of the query, search is carried out by the query subjected to the elimination of redundancy, redundancy keywords do not need to participate matching, the recall rate of search results is improved, and the search effect is improved.

The invention, which belongs to the field of artificial intelligence, relates to a robot path planning method based on a Bezier optimization genetic algorithm. With a Bezier curve, an initial solutionof a genetic algorithm and paths generated in crossing and mutation processes are optimized to eliminate a peak inflection point and reduce redundant nodes, thereby improving the path smoothness. Andthen fitness functions of a safety distance and an adaptive penalty factor are added and the path obtained by the genetic algorithm is adjusted dynamically, so that the quality of the planned path isenhanced. With the provided method, a short-distance smooth path can be searched and thus the energy losses caused by frequent running state switching of the robot due to the dramatic path turning are reduced, thereby guaranteeing the safety of the robot movement.

The invention discloses a particle swarm optimization method based on a mutual information similarity strategy. The particle swarm optimization method based on the mutual information similarity strategy is characterized by using a combined histogram method of mutual information solving to redefine a speed updating formula of h (e, g) and a particle in a combined histogram, wherein the e and the g are a path to be matched and a template path respectively; and the h (e, g) expresses generation times of a position g corresponding to a historical path at a position where an optimal path e is generated. Compared to an existing basic PSO algorithm, by using the method of the invention, iteration times are reduced, a convergence speed is increased and an average result of searching is increased too.

The invention provides a construction method of an index system, and a method and device of achieving a search. The construction method of the index system comprises the steps that a data file in need of indexing processing is obtained; the data file configuration is used for uniquely identifying the file identification of the data file, and information used for building an index is obtained from the data file; the index is built for the data file in a search engine service system according to the information used for building the data index and the file identification; the data file and the file identification are stored in a data storage service system; the search engine service system comprises multiple servers based on solid-state drives, the data storage service system comprises multiple servers based on the hard disk of a machine. The technical scheme guarantees that the index system has good search performance and the cost of achieving the index system is controlled to a certain extent.

The invention relates to the technical field of mobile robots. More specifically, the present invention relates to a grid-based contour line segment feature extraction algorithm. The method comprisesthe steps: clustering data points, sorting the data points in each cluster, segmenting the data points in each cluster, performing line segment fitting on the data points in each data segment, takingthe first three items as the method for searching the line segment support area provided by the invention, and fitting line segments in the support area by using a least square method in combination with the method. According to the algorithm, when the line segment supporting area is searched, a large memory and long calculation time are not needed, and when the line segment is fitted, the fittingmethod is simple. According to the method, the performance requirement on the rotary laser scanner can be reduced, and the influence of noisy points on the extraction effect is reduced. Moreover, parameters such as thresholds do not need to be set, the performance requirement on the rotary laser scanner is reduced, the influence of noisy points on the extraction effect is reduced, and the extracted line segment is high in resolution, robustness, efficiency and real-time performance.

The invention provides an adaptive hybrid target tracking method in a wireless sensor network, Based on particle swarm optimization and genetic algorithm, Chaotic dynamic adaptive strategy is introduced to adjust the inertia weight of particle swarm optimization algorithm, and then the optimized particle swarm optimization algorithm and double mutation adaptive genetic algorithm are combined to improve the searching effect and realize fast target tracking in WSN, especially in occlusion, noise and other disturbances. The problem of local optimal premature convergence and tracking accuracy aresolved to improve the accuracy and real-time performance of target tracking. When the infinite sensor network is used in target tracking, the target moves quickly and is disturbed by the surrounding environment, such as occlusion. The trajectory of the target is complex and diverse, with uncertainty and time-varying.

The invention relates to a Geoword-based multivariate power time series data indexing method. The Geoword-based multivariate power time series data indexing method comprises the following steps of obtaining original multivariate power time series data, and performing dimension reduction on the original multivariate power time series data; conducting Geoword encoding on dimension-reduced multivariate power time series data, and obtaining an MTSAX expression of the multivariate power time series data; constructing MTSAX data indexes of a multivariate power time series according to the MTSAX expression. Compared with the prior art, the Geoword-based multivariate power time series data indexing method can conduct dynamic division, specifies the precision, has no overlapped nodes and the like.

The invention discloses a grapefruit image segmentation method applying adaptive gauss and acoustic search. In the invention, an adaptive gauss and acoustic search algorithm is used to optimize a cluster center of a grapefruit image segmentation category. In the adaptive gauss and acoustic search algorithm, firstly, an adaptive gauss search operation operator is executed to generate a new individual; and then, according to an adaptive value of the new individual generated in a search process, a control parameter is dynamically adjusted so that search performance of the algorithm is increased. In the invention, a probability of falling into a local extremum can be reduced and segmentation precision of a grapefruit image is increased.

The embodiment of the invention provides a method and system for determining failure features in a black feature library. Features in the black feature library can be compared with sample files, when the features in the black feature library hit the sample files according with the abnormal deformation rule as source sample files of the features, it is determined that the features hit error sample files, and the features belong to failure features. The failure features are determined through the abnormal deformation rule, the used sample files can be randomly selected, and thus the failure features do not need to be searched for through a common software library any more, the failure feature search effect is improved, and safety of equipment is protected.

The invention discloses a complex network synchronization capability optimization method based on a particle swarm algorithm, and the method is characterized in that the method comprises the followingsteps: for a given complex saturated network, under the condition that the coupling strength between network nodes and the total coupling strength are limited, distributing the coupling strength between the nodes by using a particle swarm algorithm, so that the synchronization capability of the network is optimal. The invention further provides a device based on the method. By utilizing the characteristics of the particle swarm algorithm, the synchronization capability of the network can be efficiently optimized.

The invention discloses a power grid topology construction and topology analysis method and a computer readable storage medium. The method comprises the steps that a server acquires topology connection data of power grid equipment; according to the topological connection data, terminals connected with the devices are obtained respectively, connection data from the devices to the terminals are obtained and loaded to a first memory block, and the terminals connected with the devices comprise terminals of the devices and terminals of other devices connected with the terminals of the devices; according to the topological connection data, equipment connected with the terminals is obtained, connection data from the terminals to the equipment is obtained and loaded to a second memory block, and the equipment connected with the terminals comprises equipment to which the terminals belong and equipment to which other terminals connected with the terminals belong; and performing topology analysisaccording to the connection data in the first memory block and the second content block. According to the invention, the search performance of power grid topology analysis can be improved.

The embodiment of the invention provides an index structure and a retrieval method, device and equipment for unstructured data, and aims to construct the index structure for the unstructured data of aspecific business scene. The index structure comprises one of the following elements: index information generated by the system, user-defined service information of an application party and operatortrack information; classifying the unstructured data of the specific business scene based on the index structure; and according to the input keyword information, performing retrieval in the classifiedunstructured data of the specific business scene. According to the method, the retrieval performance of the unstructured data can be improved.

The invention discloses a scene information construction method and device based on an execution object. The method comprises the steps that task information input by a user is acquired; an executiontask and related task scene information are generated according to the task information; processing is performed according to the execution task execution information to generate an execution result;and the execution result is stored in the task scene information. By implementing the method and the device, the task scene information containing various information related to the to-be-executed object can be generated by fully utilizing the operation or demand of the user for the to-be-executed object (such as text), and the task scene information is applied to the subsequent operation or demand of the user, so that the information processing efficiency is improved.

The invention discloses an information extraction method for personnel information based on CNN and LSTM modeling. The method comprises the following steps: S1, acquiring personnel information; S2, constructing an LSTM-CNN neural network system; S3, classifying the personnel information, and selecting a plurality of keywords; S4, training the neural network system through the existing stored personnel information and keywords; S5, obtaining the correlation between the keywords and the personnel information through the training in the step S4; S6, inquiring different keywords for multiple times, so that the system obtains a primary retrieval result; S7,resorting the multiple initial retrieval results through the neural network system to obtain a final retrieval result; and S8, extracting and calling the final retrieval result through an operator. According to the invention, the personnel information retrieval speed is improved, and the problem that the traditional personnel informationis difficult to retrieve is solved. The neural network system can call related information in time through correlation between keywords and personnel information, and excellent search performance is shown.

The invention relates to a mobile terminal network parking method, a mobile terminal and a storage medium. The method comprises the steps that the mobile terminal determines whether the mobile terminal is in a constant-speed moving state or not according to detection data of a gyroscope and a gravity sensor of the mobile terminal, and if yes, an air pressure sensor of the mobile terminal is triggered to detect the air pressure value of the position where the mobile terminal is located; and whether the detected air pressure value is linearly changed or not is judged, and if yes, a network is searched according to a first preset rule and staying in the network. Whether the mobile terminal is in the elevator or not can be judged, the mobile phone is triggered in advance to search the networkagain, the network searching performance of the mobile terminal is improved, and a user can rapidly register to the network.