100 results about "Data diversity" patented technology

The invention discloses a mixed type processing system and method oriented to industry big data diversity application. The system comprises a distributed data collection subsystem used for collecting data from an external system, a storing and parallel calculating subsystem used for storing and calculating the collected data, and an integrated resource and system management platform used for managing the stored and calculated data. The storing and parallel calculating subsystem comprises a big data storing subsystem and a big data processing subsystem, wherein the big data processing subsystem comprises a memory calculating engine which is used for providing distributed memory abstract in a shared-nothing cluster and conducting parallel streamlined and thread lightweight processing on the collected data. The mixed type processing system and method can meet the requirements of industry big data diversity service application, the big data processing performance can be improved by ten or more times through acceleration of the memory calculating engine, and the usability, reliability and expandability of the system can be ensured through the integrated management platform.

The invention discloses mobile-crowd-sourcing-sensing-based motivation system realization method. The method comprises the following steps: determining a task model and issuing a task; determining a task value; selecting a worker according to the task demand; executing the task by the worker; and carrying out reward distribution by a utility-based multi-payment motivation mechanism. According to the invention, a mobile-crowd-sourcing-sensing-based motivation mechanism is provided and a location-based social network (LBSN) is used for task budget definition, data quality improvement, and multi-payment realization and the like; and a referable task value is provided for the task demander of the system and a utility-based multi-payment motivation mechanism scheme is also provided for maintaining data diversity and motivation fairness. With integrated utilization of the user characteristic information and task context information, the user can be motivated to participate in the task, thereby improving the participation enthusiasm of the user.

The invention provides a data augmentation method based on selection and generation and an image classification method, and the method comprises the following steps: carrying out the segmentation of an input image, and generating a plurality of image blocks to increase the number of training images; filtering the obtained image blocks, namely classifying the image blocks by utilizing a convolutional neural network, and selecting the image blocks related to the target object; image blocks obtained through filtering in the last step are reselected through multi-example learning, and image blockscontaining most areas of the object are selected; and finally, learning a corresponding relationship between the image and the text by utilizing the generative adversarial network, generating more new images by utilizing text description, and further expanding the diversity of the training image. According to the method, only one training sample and text description information thereof are used,and the image data diversity is amplified by segmenting, filtering, reselecting and generating the data. Image classification model training is carried out by using the amplified image data, thereby realizing image classification under a training sample condition.

The invention belongs to the technical field of networks, and particularly relates to a method oriented to prediction-based optimal cache placement in a content central network. The method can be used for data cache in the content central network. The method includes the steps that cache placement schemes are encoded into binary symbol strings, 1 stands for cached objects, 0 stands for non-cached objects, and an initial population is generated randomly; the profit value of each cache placement scheme is calculated, and the maximum profit value is found and stored in an array max; selection operation based on individual fitness division is conducted; crossover operation based on individual correlation is conducted; variation operation based on gene blocks is conducted; a new population, namely, a new cache placement scheme is generated; whether the array max tends to be stable or not is judged, and if the array max is stable, maximum profit cache placement is acquired. The method has the advantages that user access delay is effectively reduced, the content duplicate request rate and the network content redundancy are reduced, network data diversity is enhanced, the cache performance of the whole network is remarkably improved, and higher cache efficiency is achieved.

The invention relates to a wireless ubiquitous network system structure, belonging to the technical field of a wireless network. The wireless ubiquitous network system structure comprises a ubiquitous resource convergence layer, a ubiquitous network gateway middleware, a business capability convergence layer and a service convergence layer. The ubiquitous resource convergence layer is used for organizing isomeric ubiquitous resources and shielding the isomerism of ubiquitous resource supplying equipment; the ubiquitous network gateway middleware is used for converting 6LoWPAN (IPv6 over Low power Wireless Personal Area Networks) data output by the isomeric ubiquitous resources into IPv6 (Internet Protocol Version 6) data; the business capability convergence layer is used for distinguishing and managing the business capability of the ubiquitous resources and shielding the isomerism of wireless equipment; and the service convergence layer is used for comprehensively managing the ubiquitous resources and generating corresponding web services, and supplying a uniform application interface to a user. According to a wireless ubiquitous network system, the problem that the data of different IP (Internet Protocol) layers in a ubiquitous resource fusing process is various is solved; and a new way for the innovative development of an open and mixed ubiquitous network business is provided by a uniform application programming interface.

The invention belongs to the technical field of industrial big data application, and specifically discloses a data diversity and task diversity-based tire enterprise sales forecasting method. The method comprises five parts of: (1) theme feature extraction: extracting features of semi-structured data with respect to the tire enterprise through an LDA model; (2) structure feature extraction: mapping enterprises in the tire industry into a heterogeneous network according to a related relation and extracting structure features; (3) network structure extension: aiming at the problem of network sparsity, fusing the theme features and the structure features, extending a relationship map among the enterprises and finding potential association factors; (4) solving defects of a high-dimensional sample space by using a multi-task learning method of an LASSO method; and (5) factor feature extraction and time sequence analysis: effectively integrating and utilizing a plurality of data sets (tire demand plans, main engine plant product sales, tire sales and delivery data and the like) related to sales, and carrying out experimental data verification to improve the tire sales forecasting correctness.

To provide a data transmission method using Nt transmission antennas over a communications channel provided with Ni inputs as its feature. This method includes a bit-interleave step. The bit-interleave step has bit multiple separation DMXS, sequence-interleave SRIS, sequence storage / segmentation, segment sorting SPS, and sequence allocation ALS to form groups G1 and G2 for Nt sorted bit sequences Psqj for the purpose of transmission performed for all transmission antennas through Nt channels allocated to the same bit rank. Thereby, a basically uniformized distribution for encoded data bits is secured extending over different channel inputs, so that the method becomes suitable for data diversity.

The invention provides a baseband signal processor supporting a multi-chip architecture. The baseband signal processor comprises a main control unit, a radio frequency interface unit, a data diversity unit for the collection, packaging, unpacking and distribution of data exchange, control parameters and feedback parameters among all baseband signal processors, a chip interconnection interface unit for the high-speed data exchange among the baseband signal processors and necessary interaction of handshake signals, a symbol-level processing unit, an FFT (Fast Fourier Transform) correcting unit for correcting blocked FFT data based on an FFT time extraction algorithm to obtain high-order fast Fourier transform results, a code block-level processing unit and a memory unit. The invention further provides a processing method of the baseband signal processor. Based on the structure of the traditional baseband signal processor, part of processing units are added, so that the high-bandwidth baseband signal processor can be established by adopting the multi-chip architecture under the condition of increasing less processing resource and processing cost.

The invention discloses a data layered generation method based on fuzzy testing of an industrial control protocol. The data layered generation method comprises the following steps of: (1), performing data layering according to a network protocol or a service requirement; (2), respectively loading script configuration files in various levels; (3), analyzing the script configuration files, and generating variables in different data types; (4), sequentially generating fuzzy testing data according to the increasing level sequence of the script configuration files; (5), stitching data generated in all the levels, so that a complete data packet is formed; and (6), sending the packaged data packet to tested equipment. The data layered generation method disclosed by the invention has the advantages that: a protocol is decomposed into multiple layers; the protocol data generation complexity is reduced; protocol data in each layer can correspond to multiple script configuration files; therefore, supports are provided for data diversity; the script configuration files can be freely combined; therefore, more service requirements can be satisfied; repetitive workloads are unnecessary; and accumulation of a fuzzy testing database is facilitated.

The invention discloses a method and a system for processing crowd-sourcing multimode data. The method includes: extracting textual characteristics of textual data and image characteristics of image data from social network crowd-sourcing data, and clustering the image data based on the image characteristics; performing Hash coding on the basis of the extracted textual characteristics and image characteristics; finally, calculating Hamming distances between textual data Hash codes and image data Hash codes, and adding image data, smallest in Hamming distance, in each type of images into an optimal data set; according to the textual data in crowd-sourcing data, optimizing the image data in the crowd-sourcing data by combination of clustering and cross-modal data association. Therefore, data diversity is improved while data semantic relevance is guaranteed.

The embodiment of the invention discloses a data processing method, a router and an NDN system. The data processing method can include the steps of obtaining priority attributes of data when the data are received, setting life cycle attributes for the data according to the corresponding relation of the priority attributes and the life cycle attributes, and storing the data with the life cycle attributes in a local cache. According to the data processing method, the router and the NDN system, the life cycle attributes of the data are set according to the priority attributes of the data, the differential cache strategy can be achieved, the data diversity of the NDN system is improved, and meanwhile the data hit rate of the NDN system is improved.

The present invention discloses a real-time power load forecasting method based on an integrated network of an incremental transfinite vector regression machine, comprising an online learning stage and an online load forecasting stage. The online learning stage comprises the following steps: acquiring a first batch of power load data and influence factor data in real time, and normalizing the data; initializing an II-ESVR model; and acquiring a (k+1)th (k is not less than 1) batch of power load data and influence factor data in real time, normalizing the data, and performing incremental learning and training. The online load forecasting stage comprises the following steps: acquiring data of a batch of related influence factors in real time, normalizing the data and using the data as an input of the model; and calculating forecasting results based on the II-ESVR model and parameters of the learning stage in real time. According to the present invention, the real-time power load forecasting method based on the integrated network of the incremental transfinite vector regression machine solves the problem of instability caused by the diversity of the data, has the characteristics of "fastness, stability and accuracy", and can meet the requirements of future development of power load forecasting.

The present invention relates to a method for transmitting data including a bit interleaving step for permutating encoded data bits Tb before their transmission in the form of modulation symbols over Nt transmitting antennas. According to the invention, the bit interleaving step involves bit demultiplexing DMXS for extracting bit sequences Bsj (for j=0 to Nt) from the encoded data bits Tb, sequence interleaving SRIS for permutating the bits of each sequence Bsj (for j=0 to Nt), sequence storing and segmenting SQSS for dividing each interleaved sequence Isqj (for j=0 to Nt) into successive segments Psgj, and segment shifting SGS for applying circular shifts to simultaneous different segments Psgj (for j=1 to Nt). The invention enables to ensure that consecutive encoded bits see a maximum number of different communication channel states, and thus favours data diversity.

The invention discloses an NER-oriented Chinese clinical text data enhancement method and device. Unlabeled data and labeled data subjected to label linearization processing are obtained through data preprocessing. Unlabeled data is used, part of information in a text is masked, the masked part is predicted on the basis of reserved information, meanwhile, an entity word level judgment task is introduced, and fragment-based language model pre-training is carried out; a plurality of decoding mechanisms are introduced in a fine tuning stage, a relationship between a text vector and text data is obtained based on a pre-trained fragment-based language model, linearized data with entity labels are converted into the text vector, text generation is carried out through forward decoding and reverse decoding in a prediction stage of a text generation model, and the text generation model is used for generating text data. The method comprises the following steps of: analyzing a label to obtain enhanced data with annotation information; according to the method, the data diversity is further improved, and the quality of the enhanced data is also improved, so that the model can generate more high-quality enhanced data.

The invention discloses an MR image segmentation method and device based on a conditional generative adversarial network, and the method comprises the steps: training a cGAN network through a low-level MR image y and a segmentation mask x, and obtaining a trained cGAN model; automatically generating an artificial image y' by using the trained cGAN model according to the input segmentation mask x;pre-training a U-net network by utilizing the artificial image y' and the segmentation mask x; continuously training the U-net network through the low-level MR image y and the segmentation mask x to obtain a trained U-net model; and performing MR image segmentation by using the trained U-net model. The invention also provides a device for realizing the method, terminal equipment and a computer readable storage medium, the data diversity of an original data set is expanded and enriched, an image segmentation result is closer to a real image, and the convergence speed, the loss and the segmentation precision in a segmentation network training process are higher.

The invention discloses a data diversity combining method and device, communication equipment and a storage medium, and the method comprises the steps: selecting and reserving data for diversity combining according to the subcarrier position information of narrow-band noise and a preset selection strategy; generating a diversity combining coefficient according to the reserved data; and according to the diversity combining coefficient, carrying out diversity combining on each path of data. According to the embodiment of the invention, the data diversity combining method and device, the communication equipment and the storage medium are applied to a power line communication system, can adapt to a power line communication environment, and select and reserve data with less narrow-band noise pollution for adaptive diversity by utilizing the position relationship between the data and the narrow-band noise, thereby reducing the influence of the narrow-band noise on signals, obtaining better diversity gain, and improving the reliability of data transmission.

The invention relates to a data reconstitution short-term traffic flow prediction method used for solving the problem of traffic jam. The method includes: acquiring traffic flow data on working days, nonworking days and important festivals so as to guarantee data diversity; providing a data reconstitution method, reconstituting the acquired traffic flow data in the same continuous time period to natural numbers, creating a short-term traffic flow database, and updating the short-term traffic flow database regularly to adapt to traffic flow changes; reconstituting the traffic flow data of the current continuous time period to natural numbers; comparing the current reconstituted traffic flow data with historical data in the short-term traffic flow database in size, and capturing a similar sample range; providing a deviation value calculation method to obtain deviation values of all samples in the similar sample range; predicating by utilizing two samples with the minimum deviation value; reconstituting mass historical traffic flow data to enable the similar samples to be searched more simply, quickly and accurately. The traffic flows on the working days, nonworking days and important festivals need not to be predicated separately, and predication rapidness and accuracy are improved.

The invention discloses a report form generation method and system based on big-data template technology. The method includes: carrying out format conversion on imported original data, and saving the data into a database; synchronizing the data in the database to a big-data cluster; carrying out data analysis and mining processing on the data in the big-data cluster to obtain report form data; and generating and obtaining a report form according to the report form data and a required report template. The system includes a data collection module, a data synchronization module, a data analysis and mining module and a report form generation module. According to the method and system, multi-format data can be converted into data of a unified format through an adapter, the exported report form is consistent with a defined report form style and format thereof, compatibility with an original data statistics method and a big data-based data mining method are also provided, report form types and data diversity and depth are enriched, and important support is provided for enterprise management and decision-making. The method and system can be widely applied to the technical field of data report forms.

The invention relates to a load identification method for a baseline load model based on a Bayesian classification method, and the method comprises the following steps: 1, determining whether any twogiven user characteristic quantities are in statistical correlation or not, and eliminating redundant characteristic quantities; 2, forming a training set by the user characteristic quantities which are subjected to judgment and statistics correlation and are subjected to redundant characteristic quantity elimination; 3, calculating a prior probability for each data category in the training set, and further obtaining a corresponding Bayesian probability; 4, calculating the baseline load of the prediction day by combining the Bayesian probability with the power load of the prediction day; and 5, identifying the peak shifting potential load by utilizing the obtained baseline load. Compared with the prior art, the method has the advantages that the single predicted daily baseline load is avoided, the peak shifting potential load cannot be identified, the data diversity is high, and the like.

The invention discloses an object detection model training method and device, an object detection model detection method and device, equipment and a medium. The object detection model training methodcomprises the steps of obtaining an initial network model, wherein the initial network model comprises at least two detection sub-modules for objects in different scale intervals; adjusting a corresponding data augmentation strategy according to the scale interval; and augmenting the training sample images by using the adjusted data augmentation strategy, and training the corresponding detection sub-modules in the initial network model by using the augmented training sample images according to the scale interval to obtain an object detection model. According to the method, the network complexity does not need to be increased, the data diversity is increased by adjusting the data augmentation strategy, and the detection rate of the object detection model obtained by training can be effectively improved.

The present invention relates to a method for transmitting data between a transmitter provided with at least two transmitting antennas (ta1, ta2 . . . taNt) and a receiver. The method according to the invention further includes a symbol spreading step in the course of which components Zij (for i=I to S and j=I to Nt) of a predetermined number of successive symbols ZI . . . ZS are to be spread over time before being transmitted over said transmitting antennas (ta1, ta2 . . . taNt). The data transmitted at any given moment will thus not be representative of a single symbol Zi (for i=1 to S), as is the case in known MIMO systems, but will represent a mixture between Nt components of S successive symbols, which introduces data diversity with respect to time as perceived at the receiver end.

The invention discloses a GAN-based data enhancement unsupervised trademark retrieval system. The unsupervised trademark retrieval system comprises a GAN data enhancement module, an instance distinguishing module and a trademark retrieval module. The GAN data enhancement module is used for generating a trademark data set and expanding a training set; the instance distinguishing module is used fortraining an unsupervised network and extracting trademark features; and the trademark retrieval module is used for calculating the similarity between the trademark database and the trademark featuresto be retrieved and sorting the trademark features according to sizes. First, a trademark data set is used to train an adversarial generative network. Then, the trained GAN module is used for generating an enhanced data set, and an original trademark data set is added to form a new trademark data set; and finally, applying the new trademark data set to a training instance distinguishing module. Inthe trademark retrieval module, trademark features of a trademark image to be retrieved and a new trademark data set are extracted through a trained instance distinguishing module. According to the method, the problems of difficult data labeling and insufficient data diversity in trademark retrieval are effectively solved.

The invention discloses a Spark-based diversity graph sorting method for large-scale graph data. The diversity graph sorting of the graph data is carried out in combination with a classic personalized Page Rank algorithm and a distance-based diversity measurement method by taking the diversity graph sorting of the large-scale graph data as a goal and taking a method for measuring a distance between nodes in the graph data as a basis. The method has the advantages of expandability, higher efficiency and the like, meets the data storage and calculation requirements of the diversity graph sorting of the massive graph data, and provides a powerful technical support for to-be-solved key problems in massive graph data analysis processing and mining, and the like.

The invention discloses a training method for improving robustness of a multi-branch prediction single model in an Internet of Things scene, and the method comprises: segmenting training data under the structure of a cloud terminal, distributing the training data to terminals, forming an environment with data diversity, setting different optimization targets for each terminal, and improving the structural diversity. And each terminal updates the weight issued by the cloud by depending on the training data and the optimization target of the terminal, then the cloud recovers the weight of each terminal for aggregation, finally training is completed through multiple cycles, and the optimal weight of the multi-branch prediction single model is stored. According to the method, the structural diversity of the multi-branch prediction single model is successfully combined with the data diversity under the scene of the Internet of Things, so that the robustness of the multi-branch prediction single model is further improved, the classification capability of adversarial samples is improved, and the multi-branch prediction single model can safely and reliably complete a classification task.

The invention relates to an incremental data enhancement method for visual question and answer model training, and the method comprises the steps: obtaining an original training data set, wherein theform of a training sample in the data set is (images, texts and answers), and the texts are formed by a natural language sequence; obtaining sentence length distribution of a natural language sequencein the original training data set and word frequency distribution of each word, and determining a minimum sentence length threshold value and a maximum sentence length threshold value based on the sentence length distribution; and expanding a natural language sequence in the training sample according to the minimum sentence length threshold, the maximum sentence length threshold and the word frequency distribution to achieve data enhancement. Compared with the prior art, the method has the advantages of data diversity, good efficiency, simplicity and the like.

The invention discloses a large-screen display configuration method. The large-screen display configuration method comprises three stages as follows: display verification, display configuration and display loading, wherein the display verification refers to verification of types, page elements, technical means and resolutions during large-screen display; the display configuration refers to configuration of the types, resources, page element types corresponding to the resources, the used technical means and the resolutions; the display loading refers to loading of configured information on a large screen in a graphic form through the types, the resources and element data. Compared with the prior art, the large-screen display configuration method has the advantages as follows: the problems of data diversity (tables, graphics, other system pages, maps, texts, etc.), management fragmentation, and too high coupling of a presentation layer and a development layer in the conventional large-screen display are solved, high cohesion and low coupling are achieved, independence between a foreground page and a background technology is improved, and the large-screen display configuration method is high in practicality, wide in application range and easy to popularize.

The invention provides a health record data generation method and device. The method comprises the following steps: training a plurality of groups of real health archive data in an InfoGAN architecture to obtain an archive recognizer; training in an InfoGAN architecture through the target disease type set and the target health level set to obtain an archive generator; the archive recognizer and the archive generator are alternately subjected to optimization training through a loss function of an InfoGAN architecture; data authenticity discrimination is carried out based on the current archiverecognizer and the current archive generator after each optimization training; a mode until the judged real probability value is close to a preset probability value is adopted; Therefore, the file generator is used as a target file generation model to simulate and synthesize health file data with high authenticity and high data accuracy, and the data diversity is ensured through the respective composition of the target sick type set and the target health level set, so that the medical research is facilitated.

The invention discloses a user satisfaction analysis method and device. The user satisfaction analysis method includes the following steps: integrating data of a BSS domain and an OSS domain, and obtaining indicator data for predicting the user satisfaction; cleaning and marking the indicator data; and predicting the user satisfaction by using an improved random forest algorithm based on the cleaned and marked indicator data. According to the scheme of the invention, the OSS domain and the BSS domain are subjected to fusion analysis, the data diversity can be improved, multi-dimensional data can be formed, and by analyzing a balanced random forest algorithm that is used to improve multi-dimensional indicator data, the accuracy rate, recall ratio and precision ratio of predicting the user satisfaction can be effectively improved.

The invention provides an abnormal image detection method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining an abnormal image sample, the abnormal image sample being an image obtained by rendering a second application package, and the second application package being a package obtained after adding a problem code to a first application package; obtaining a normal image sample, wherein the normal image sample is an image obtained by rendering the first application package body; based on the abnormal image sample and the normal image sample, the initial detection model is trained, a trained target detection model is obtained, and the target detection model is used for outputting the probability that the input image is a normal image or an abnormal image. According to the method, the data size and data diversity of the abnormal image sample are improved, so that the accuracy of detecting the abnormal image by the target detection model is improved.

The invention belongs to the technical field of computer graphics, and relates to a three-dimensional human body attitude estimation method, which can accurately regress to obtain a three-dimensional posture only through a two-dimensional posture obtained by a single image, gets rid of the technical constraint that the accurate three-dimensional posture can be obtained by depending on high-cost hardware, and improves the accuracy of human body attitude estimation. According to the method, the cost of human body three-dimensional posture-dependent applications such as human-computer interaction, augmented reality and virtual reality is greatly reduced. Meanwhile, the three-dimensional posture capturing precision of unusual actions is greatly expanded, more complex actions can appear in the applications such as human-computer interaction, and the interestingness and practicability of the applications are greatly improved; the method is scientific and reliable in principle, data diversity is expanded on the basis of an existing data set, the model generalization ability is improved, vivid and natural three-dimensional human body postures are obtained by relying on image data collected by a monocular camera in a richer real scene, the precision of three-dimensional attitude estimation of unusual actions can be remarkably improved, and the method is suitable for popularization and application. The method can be applied to more diversified scenes.