42 results about "Data synthesis" patented technology

This invention relates to the field of knowledge base construction technology, specifically disclosing a method and system for constructing an industry knowledge base based on text-based data synthesis. The method includes entity recognition of multi-source text data, constructing a knowledge point sequence, and marking key knowledge points in the knowledge point sequence; expanding the key knowledge points based on a large language model to generate an expanded knowledge point set; constructing a knowledge graph based on the expanded knowledge point set, and constructing a preset number of knowledge connection paths in the knowledge graph; classifying the knowledge connection paths and inserting them as knowledge skeletons into the industry knowledge base. When dealing with massive amounts of multi-source text data, this invention expands the keywords based on a large language model to obtain expanded knowledge points, then constructs a knowledge graph, extracts node paths from the knowledge graph, uses the node paths as knowledge skeletons, classifies them, and stores them in the knowledge base. While the resulting knowledge base still contains a large amount of information, it significantly simplifies its size.

In an example embodiment, generative artificial intelligence (GAI) is used to synthesize data to be used to train and / or validate a segmentation model. A GAI model may be fine tuned with real training data, but then may be able to generate its own data. The data is image data, and the GAI model may use a real background image and then generate a fake defect and insert the fake defect in the proper place in the real background image.

This application provides a method, apparatus, equipment, and storage medium for efficient on-site quality control of node data. The method includes: selecting target node synthetic seismic data for a target work area; performing maximum amplitude limitation processing on the target node synthetic seismic data to obtain node synthetic seismic data to be processed, wherein the recording length of the target node synthetic seismic data is within a preset time period and the number of traces is within a preset range; for each shot's node synthetic seismic data to be processed, determining whether the current shot's synthetic seismic data is an empty file based on the amplitude of the sampling points in each trace of the current shot's synthetic seismic data, and removing the empty files; for the synthetic seismic data after removing the empty files, statistically analyzing the synthetic quality and the working status of nodes in multiple domains based on the non-working trace head attributes and the multi-domain overlay profile trace head attributes, and performing high-quality node data synthesis checks and node status quality control for massive 3D node data.

Systems and methods include determination of a plurality of event types, determination of a plurality of associated metrics for each of the plurality of event types, generation of synthetic time-series data of each of the metrics associated with the event types for each of the plurality of event type, the synthetic time-series data representing metric anomalies and events of the event type, training of a first system to generate anomaly values based on the synthetic time-series data, training of a second system to classify event types based on the anomaly values, determination of an anomaly value for each data instance of actual time-series data, determination of event types for each determined value, selection of data instances based on the determined values and event types, reception of an event type for each selected data instance, and re-training of the second system based on the selected data instances and event types.

The application discloses an intelligent evaluation system and method for a knowledge base question answering application, and relates to the technical fields of artificial intelligence, natural language processing and multi-agent collaborative system. The system comprises a data preprocessing module, a data synthesis module, a data screening module, a data scoring module and an adaptive weight adjustment module. The data preprocessing module performs data preprocessing on enterprise original documents and generates a data knowledge base. The data synthesis module repeatedly asks and answers each paragraph of the data knowledge base by using a large language model to generate a question and answer pair. The data screening module screens out high-quality samples through screening agent collaboration and a MACA framework mechanism. The data scoring module scores the question and answer pair through scoring agent collaboration and a MACA framework mechanism. The adaptive weight adjustment module dynamically updates the weight of each screening and scoring dimension by using an exponentially weighted moving average algorithm.

The application discloses a kind of dual-hand dexterous grasping data synthesis and grasping generation method, system and robot for multi-size object.Data synthesis method is: local grasping area is constructed on object surface, target area pair is determined and posture is initialized after geometric screening and mechanical stability evaluation;Decoupling optimization objective function including local force closure, contact distance and area consistency term is constructed, and left and right hand postures are decoupled and optimized;Feasible sample is screened and motion trajectory is generated by collision detection and inverse kinematics solution, and large-scale data set is constructed.Grasping generation method is: based on object point cloud cooperative prediction master-slave grasping perspective;Scale self-adaptive grasping anchor point is determined;Local features and global features are fused to construct conditional features;Generation model is used to generate dual-hand coordinated grasping posture in the form of pose increment relative to anchor point.The application significantly improves data generation efficiency and quality, and enhances the generalization ability of model to multi-size object and the executability of grasping.

Method for storing validation data (23) in a data storage system (12), wherein the validation data (23) are intended for validating systems for processing large amounts of data; wherein the data storage system (12) is configured to provide the validation data (23) for validating at least one system for processing large amounts of data; and wherein the validation data (23) comprise at least some data recorded by sensors (2); comprising the following steps: a) receiving the validation data; b) determining (7) at least one statistical property of the received validation data; c) performing compression (6) of the validation data to produce compressed validation data (11) using a lossy compression method; d) determining at least one statistical comparison property;e) Performing a comparison of the at least one statistical property with the at least one statistical comparison property, wherein the comparison checks whether a deviation between synthesized validation data (20) producible on the basis of the compressed validation data and the validation data received in step a) lies within a predetermined tolerance range, wherein the tolerance range is chosen such that deviations are detected which lie within a signal noise caused by properties of at least one sensor (2) with which validation data were recorded, wherein the tolerance range is further chosen such that the distance between validation data (20) synthesized on the basis of the compressed validation data (11) and the recorded reality is not greater than the distance between the validation data received in step a) from the same reality;and f) Storing the compressed validation data (11) in the data storage system (12) if, in step e), it was determined that the deviation is within the specified tolerance range.;

The application discloses a table synthesis method, device, equipment and medium, and belongs to the field of artificial intelligence. The method comprises the following steps: based on an original table, data synthesis strategies in a strategy pool, a first large model and a second large model, performing multi-round iterative synthesis processing to obtain a plurality of synthesized tables, each round of iterative synthesis processing corresponding to at least one synthesized table, one synthesized table corresponding to one data synthesis strategy, and a data synthesis strategy corresponding to an (i+1)th round of iterative synthesis processing being obtained by variation of a data synthesis strategy corresponding to an ith round of iterative synthesis processing, each data synthesis strategy comprising at least one of a field shunt mapping rule, a code prompt word and data generation code; obtaining evaluation scores of each synthesized table, and determining M1 synthesized tables with the highest evaluation scores as target synthesized tables. According to the embodiment of the application, the quality and effect of the synthesized table can be improved.

The application discloses a crop phenological season detection method based on stationary satellite data and belongs to the technical field of remote sensing monitoring. The method comprises the following steps: determining a research area range and a time interval, acquiring Himawari-8 AHI data, and obtaining a limited VI time sequence through band calculation; performing cloud identification on the AHI limited VI time sequence through a fast cloud detection method to obtain an AHI cloud-removed VI time sequence; performing data synthesis on the AHI cloud-removed VI time sequence through an n-day data 90th percentile synthesis method to generate an AHI n-day VI time sequence, and then smoothing the VI time sequence by using an SG filter; and performing phenological season detection on the smoothed AHI n-day VI time sequence by using an improved SMF-S method. The method can ensure the accuracy of the season detection, greatly reduce the data quantity of the early-stage processing, and is superior to the season detection accuracy of polar-orbit satellite MODIS data with the same spatial resolution.

The application provides a software development data synthesis method and device based on code repository evolution process, a medium and a program product. The method comprises the following steps: obtaining development process data with code merging requests in a code repository, and structuring development records; determining the evolution relationship between the code merging requests based on the structured development records, and constructing a code repository evolution chain; mapping the structured development records in the code repository evolution chain to staged tasks; and executing corresponding stage task inputs on the repository versions of the staged tasks to generate candidate execution tracks; evaluating the generated patches in the candidate execution tracks based on real patches, and filtering the candidate execution tracks based on stage evaluation scores, and converting the candidate execution tracks meeting the filtering conditions to a target data format to generate software development data. The application improves the reproducibility, verifiability and quality controllability of the software development data, and makes the software development data closer to the real software development scene.

This application provides a communication method and apparatus, relating to the field of communication, which can solve the problem of insufficient effective data in wireless communication networks. In this method, a data synthesis function can obtain data for generating synthetic data from a first device according to a first message sent by a control function to trigger the generation of synthetic data, thereby generating first synthetic data. This allows synthetic data to be generated based on existing data in the wireless communication network, expanding the effective dataset in the wireless communication network for model training, inference, fine-tuning, etc.

The application discloses an X-ray image denoising method based on physical driving data synthesis and local constraint, and particularly relates to the field of X-ray imaging denoising technology, and comprises the following steps: collecting an X-ray image through an imaging system; introducing a clean signal Isg, background texture Itx, particle cluster noise Icl and a calibration offset Offset, and superimposing them to the X-ray image to generate a synthetic noise image; performing domain adaptive enhancement and double-path denoising, and mapping the synthetic noise image to obtain an enhanced noise image; through a synthetic noise modeling process based on a physical imaging mechanism, the application uniformly models and superimposes a hot spot radiation form, background irradiation fluctuation and particle cluster pulse noise to generate a synthetic noise image, so that the training data is highly consistent with real X-ray experimental data in terms of statistical distribution and physical characteristics, and the adaptability and stability of the double-path denoising model in a complex scene are improved.

This invention discloses a method for synthesizing plant disease image data based on a lesion information transmission and diffusion model, belonging to the field of plant disease detection technology. The method first detects and segments lesion regions in the original disease image using a fine-tuned visual basic model. Second, a connectivity graph is constructed based on the spatial relationships of lesion locations, and a mask image is generated that preserves the lesions and obscures healthy tissue, along with corresponding related images around the lesions. Then, the mask image is used as input to a diffusion model, and semantic features extracted from related images are used to guide the diffusion process to generate a high-fidelity disease image. Finally, using the trained model, a novel, realistic plant disease image is synthesized in a designated area of ​​a healthy leaf image. This invention can accurately capture the lesion features of plant diseases and generate high-quality disease images, providing high-quality training data for disease diagnosis models.

The information processing device is an information processing device for generating driving assistance information for a vehicle. An information processing device includes: a detection data obtainer that obtains detection data, the detection data being three-dimensional data obtained by detecting an object located outside the vehicle; a data complementation that extracts object data from a plurality of pieces of three-dimensional data of objects stored in advance, the object data being three-dimensional data of the object corresponding to the detection data, and complements the detection data with the object data to generate complementary data that is three-dimensional data; and a data synthesizer that synthesizes the complementary data and vehicle data, the vehicle data being three-dimensional data of the vehicle, to generate synthesized data that is three-dimensional data including the vehicle and an external area of the vehicle.

The application provides a roadside weak light data synthesis method and device for simulating imaging characteristics of a sensor, and relates to the technical field of computer vision and intelligent transportation, and comprises the following steps: acquiring a daytime roadside image and performing normalization processing; simulating a nonlinear brightness response of the sensor by exposure scaling and Gamma correction; applying a spectral scaling coefficient to simulate spectral shift of a night artificial light source; constructing a dark corner intensity mask to simulate lens optical defects; performing adaptive smoothing on a dark area in a color space to simulate texture loss; establishing a Poisson-Gaussian mixed model to simulate physical noise of the sensor; performing ISP fine tuning and quantization to synthesize a weak light image. In view of the problems that weak light data is scarce in a roadside sensing scene and a traditional image darkening method cannot reflect real physical characteristics, the weak light data generated by the method is consistent with a real night scene in terms of pixel visibility distribution, texture structure reservation and color noise characteristics, and the robustness and detection accuracy of a roadside all-weather traffic sensing algorithm are effectively improved.

The application discloses a local differential privacy graph data synthesis algorithm based on a neighborhood structure, and comprises the following steps: obtaining the disturbed node degree and the adjacency list of each user, and calculating a projection threshold value for constraining the sensitivity of triangle counting; using the projection threshold value and the disturbance mechanism to guide the user to calculate and report the core neighborhood structure information, and jointly optimizing the collected two neighborhood statistics of the degree and the triangle count value, correcting the structural inconsistency, and completing the generation of the target statistics; using the corrected target statistics to group all nodes, and sequentially executing an edge generation algorithm within and between groups, preferentially meeting the triangle count target, and then meeting the degree value requirement, and completing the construction of the synthesized graph meeting the local differential privacy protection. Under strict privacy protection, the application significantly improves the availability and fidelity of the synthesized graph data by maintaining the key neighborhood structure of the graph.

The invention provides an explosive source node data synthesis and quality control method and system based on a shared memory. The method comprises the following steps: step 1, constructing a node data information table; step 2, constructing an observation system data table; step 3, creating a shared memory space; 4, cutting node data, and synthesizing a detection line record; 5, calculating to obtain a detection line record quality analysis result according to the detection line record; step 6, according to the detection line record, performing batch merging according to a field file number sequence to obtain all single shot records; and step 7, according to the synthesized single-shot record, calculating to obtain a single-shot quality analysis result. According to the explosive source node data synthesis and quality control method and system based on the shared memory, the problems that single-shot records are directly cut and synthesized one by one and quality control cannot be carried out in time in the past are solved, the cutting synthesis efficiency of the explosive source node acquisition data is greatly improved, and the quality of field seismic acquisition data can be monitored in time.

This application provides a method, apparatus, storage medium, and processor for processing insurance regulatory reporting data, belonging to the field of big data technology. The method includes: configuring reporting nodes based on the reporting party's organizational structure and asset status; configuring report synthesis rules based on report types; collecting source data from different custodian bank systems according to the current target report type; verifying the source data according to the current target report type to obtain data to be reported; and synthesizing the data to be reported into a target reporting report according to the report synthesis rules corresponding to the current target report type and the reporting nodes. Configuring reporting hierarchy nodes based on the reporting party's organizational structure allows for flexible adaptation to different reporting hierarchy requirements. Configuring report synthesis rules based on report types automatically acquires, verifies, and summarizes the reported report data level by level, achieving adaptation to different insurance company reporting hierarchy management. Automated report synthesis according to preset rules improves reporting efficiency.

This application provides a multimodal medical data synthesis method, apparatus, electronic device, and storage medium, belonging to the field of artificial intelligence technology. The method includes: acquiring medical image-text pairs; determining a first prompt word based on the medical role type, medical question meta-attributes, and medical image-text pairs of a first intelligent agent; inputting the first prompt word into the first intelligent agent to obtain medical question data output by the first intelligent agent; determining a second prompt word based on the medical question data and medical image-text pairs; and inputting the second prompt word into a second intelligent agent to obtain medical image question-and-answer pairs output by the second intelligent agent. This application can generate medical image question-and-answer pairs with multi-role, multi-scenario, and multi-dimensional professional questions based on the same medical image-text pairs, under different medical roles and different medical question meta-attributes, meeting the demand for high-quality training data for medical image question-and-answer model training.

The application discloses a trajectory data synthesis method based on a diffusion probability model and dynamic simulation application thereof. The method aims to solve the problem that existing trajectory generation models lack diversity, authenticity and are difficult to effectively integrate complex environmental constraints. The method models the trajectory synthesis problem as a conditional denoising process, gradually converts real trajectories into pure noise through a forward diffusion process, and trains a conditional denoising network to learn the reverse process, thereby recovering realistic and environment-constrained future trajectories from noise. In particular, the application introduces an environment perception guiding mechanism, which integrates high-dimensional map information into the sampling process of the diffusion model, ensuring that the synthesized trajectories meet the physical feasibility and non-collision in the dynamic simulation environment. The method can generate high-precision, high-diversity and environment-consistent synthesized trajectory data, greatly improving the efficiency of dynamic simulation and algorithm testing in the fields of autonomous driving and urban traffic planning.

The application discloses a network anomaly diagnosis system, an operation and maintenance management device and a network diagnosis method thereof. The network diagnosis method uses a network diagnosis model to detect log data. The network diagnosis model is trained by using a log data set provided by a GAN neural network model trained by a method for training the GAN neural network model. The method for training the GAN neural network model comprises the following steps: acquiring data; performing data enhancement on the data by using the GAN neural network model, and outputting virtual data; selecting important data with important features from the virtual data, and performing data synthesis on the important data by using a loss function, so as to realize sample enhancement of the important data.

A method (1000) includes obtaining (1002) a dataset (114, 302) and identifying (1006) a sensitivity level (310) associated with the dataset. The method also includes dynamically selecting (1008), based on the sensitivity level and additional information (304) associated with the dataset, at least one of: (i) a scanning technique (314) that defines how the dataset is scanned for sensitive information to be removed, (ii) a masking technique (316) that defines how the sensitive information is to be removed, and (iii) a data synthesis technique (318) that defines how the sensitive information is to be replaced with synthetic data. In addition, the method includes applying (1010) the at least one of the scanning technique, the masking technique, and the data synthesis technique during generation of an anonymized dataset (332), where the anonymized dataset lacks the sensitive information.

This invention provides a method for testing the scattering of electrically large targets based on component decomposition, belonging to the field of electromagnetic scattering measurement. The method includes: dividing the complete target into multiple independent sub-components; covering the cross-section of each independent sub-component with absorbing material; performing scattering tests on each independent sub-component; filtering out the scattering components from the cross-sections during component testing; performing phase correction on the test data of each component; and employing component scattering data synthesis technology to obtain the complete target scattering data. This invention can achieve higher testing accuracy, faster testing speed, and better testing flexibility, significantly alleviating the limitations of testing sites for electrically large targets and improving the testing capabilities for electrically large targets.

This invention provides a multi-domain adaptive target tracking method and system based on domain-customized adaptation. The method includes a data synthesis stage and a multi-weather general adaptation stage. In the data synthesis stage, a preset text diffusion model is used to diffuse and output multiple target domain search frames. The multi-weather general adaptation stage includes the following steps: constructing a first set of multiple target domain search frames, constructing a second set of multiple target domain search frames and source domain search frames, and extracting two frames corresponding to the same source domain search frame from the first set and the second set respectively; inputting the frames extracted from the first set into a preset teacher network, and the teacher network outputs a first localization score map; inputting the frames extracted from the second set into a preset student network, and the student network outputs a second localization score map; calculating a first loss function based on the first localization score map and the second localization score map, and training a student model and a teacher model based on the first loss function.

The application discloses a model inference method and device, equipment and storage medium, relates to the technical field of large model application, and comprises the following steps: obtaining regular training data suitable for an adaptive data synthesis scene, a target model to be deployed and a draft model; performing parallel training on the draft model according to the regular training data and a preset parallel decoding strategy to obtain a parallel draft model; modifying the source code of an inference framework suitable for a domestic graphics card, inserting parallel draft decoding logic, and obtaining a customized inference framework supporting parallel draft inference; and based on the customized inference framework, jointly inferring the target model and the parallel draft model, accelerating the processing of an inference request to be sent, and completing effect verification. The application jointly infers by constructing a parallel draft model and adapting a domestic graphics card inference framework, effectively alleviates the problem of shortage of domestic graphics card computing resources, overcomes the defects of insufficient adaptation of an existing inference framework and high inference delay, and significantly improves the model inference efficiency.

The application provides a construction method of an error-related potential classification model based on a time convolution network, and the construction method comprises the following steps: acquiring an error-related potential signal sample set, calculating the difference between correct category samples and error category samples in the error-related potential signal sample set, and acquiring the number N of error category samples to be generated; using a data synthesis technology, generating N error category artificial samples based on the error category samples, and combining the error category artificial samples with the error-related potential signal sample set to form a training set; constructing an MS-ATCNet model, training the model by using the training set, and obtaining an error-related potential classification model; wherein the MS-ATCNet model comprises a discrete wavelet transform module, a convolution reconstruction module, a multi-scale attention aggregation module, a time convolution module and a classifier module.

The application provides an electroencephalogram data synthesis method, device and equipment and a storage medium, and belongs to the field of biological signals. The method comprises the following steps: in response to an electroencephalogram data synthesis task, loading corresponding head model information, wherein the head model information is biophysical basic information; synthesizing electroencephalogram scalp signals according to the head model information and the electroencephalogram data synthesis task to obtain scalp signal data, and outputting electroencephalogram signal causality during the synthesis of the electroencephalogram scalp signals to obtain causality feature data; generating simulated noise for electroencephalogram extraction according to the head model information and the electroencephalogram data synthesis task to obtain noise signal data; and synthesizing electroencephalogram data according to the electroencephalogram data synthesis task, the scalp signal data, the noise signal data and the causality feature data. By adding the electroencephalogram signal causality and the simulated noise for electroencephalogram extraction into the electroencephalogram data, the generated electroencephalogram data has a causal true value and is more accurate.

A system, method, and computer-readable medium for generating synthetic data are described. Improved data models for databases may be achieved by improving the quality of synthetic data upon for modeling those databases. According to some aspects, these and other benefits may be achieved by using numeric distribution information in a schema describing one or more numeric fields and, based on that schema, distribution-appropriate numerical data may be generated. The schema may be compared against actual data and the schema adjusted to more closely match the actual data. In implementation, this may be effected by storing a schema with distribution information and / or one or more parameters, generating synthetic numerical data based on the schema, and, based on a comparison with actual data, modify the schema until the synthetic data is statistically similar to the actual data. A benefit may include improved database performance and indexing based on repeatable, statistically appropriate, synthetic data.

The present application relates to a kind of data synthesis method and data synthesis system of training data and evaluation data of large language model.The method at least includes the following steps: data preprocessing step, the original data of financial payment field is preprocessed to generate real data, seed data is preprocessed and structured seed data and the variant of the structured seed data are generated;Training data synthesis step, based on the structured seed data and the variant of the structured seed data Synthesize training data;And data expansion step, expand the training data to obtain candidate training data and generate candidate evaluation data with the semantic difference of the candidate training data.According to the present application, the generalization performance of model in real financial scenario and evaluation fairness can be significantly improved.