31 results about "Sparse data sets" patented technology

The invention provides an urban built-up area air quality space-time evolution and exposure risk prediction method, and relates to the technical field of electrical digital data processing.The space-time evolution method comprises the steps that a to-be-monitored target area is discretized into a cubic sampling domain, an execution flight path of an unmanned aerial vehicle is planned, and a sparse data set containing five-dimensional features is constructed; executing iterative up-sampling operation, and converting the sparse data set into a dense feature point set covering a sampling domain; calculating a pollutant transmission flux vector of each node in the dense feature point set, constructing a weighted reverse tracking vector pointing to a suspected source, and constructing a three-dimensional traceability probability cone in combination with turbulence intensity; extracting a centroid offset coefficient and an axial dispersion, and identifying an emission working condition label of the pollution source through an emission working condition ternary discrimination model; executing multi-source probability field superposition fusion, and determining an estimation coordinate of a pollution source; and executing forward space-time diffusion simulation based on the estimated coordinates, and outputting a three-dimensional risk early warning graph.

This disclosure provides a mechanism for the enrichment of sparse datasets using language models. By training language models on the specific distribution of known values in a dataset, missing values can be predicted, and the predicted values added, thereby resulting in a more complete dataset. This method also facilitates the enhancement and augmentation of datasets by predicting values for new properties that were not previously available. The approach proves particularly effective at scale, transforming large sparse datasets into more complete and enhanced datasets. Masking language modeling may be employed to train language models capable of generating representations of technical data. Training data includes corpuses of technical data that may be represented as text strings. These pretrained models are fine-tuned to predict various properties. The resulting models can predict missing values in large technical datasets, providing valuable data for guiding scientific research.

The invention relates to the technical field of geotechnical parameter analysis, and discloses a sparse data probability distribution modeling method applied to geotechnical parameter data acquisition, which comprises the following steps of: establishing a sparse data set according to acquired sample data and sample quantity; based on the sparse data set, adopting a Gaussian function as a kernel function, and obtaining an approximate probability density function and an approximate cumulative distribution function through a kernel density estimation algorithm; obtaining an empirical cumulative distribution function of the sparse data set; obtaining a maximum absolute difference through a difference value between the approximate cumulative distribution function and the empirical cumulative distribution function; taking a plurality of random bandwidth parameters as an initial group of chromosomes, carrying out optimization solution on the maximum absolute difference through a genetic algorithm, and obtaining the chromosome corresponding to the maximum fitness as an optimal bandwidth parameter of an approximate probability density function and an approximate cumulative distribution function; and substituting the optimal bandwidth parameter into the approximate probability density function and the approximate cumulative distribution function to obtain a target approximate probability density function and a target approximate cumulative distribution function.

The invention discloses a sparse large-scale multi-objective optimization method based on a generative adversarial network, and relates to the technical field of large-scale multi-objective optimization, and the method comprises the following steps: carrying out the preprocessing of a large-scale high-dimensional sparse data set, so as to reduce the data dimension and noise, and guaranteeing that the solution space exploration of a sparse region is not ignored in the optimization process. Through sparse data preprocessing, a sparsity control mechanism and multiple rounds of adversarial learning circulation, the solution set diversity and coverage rate of the generative adversarial network in multi-objective optimization are improved, the mode collapse phenomenon is effectively avoided, and the optimization result is more comprehensive. When high-dimensional sparse data is processed, through feature extraction, dimensionality reduction and dynamic learning rate adjustment, calculation complexity is reduced, training efficiency and stability are improved, it is ensured that a generator can explore a sparse region and a high-density region of a solution space in a balanced mode, and efficient and reliable technical support is provided for a complex decision-making scene.

A system may access a set of training data and determine a timeframe associated with a positively labeled data item of the training data. A system may generate at least two new positively labeled data items based on the positively labeled data item to generate augmented training data. A system may train a machine learning model by applying the augmented training data as input to a machine learning model, and modifying a weight of the machine learning model.

The invention relates to a multi-beam sounding data quality real-time evaluation method based on a Delaunay triangulation network model, and the method comprises the steps: carrying out the thinning of an original discrete data set, generating a thinned data set, and constructing a Delaunay triangulation network; performing fragmentation processing on the original data set according to the Delaunay triangulation network, and extracting an accurate data range of the fragmented data set; and respectively carrying out gridding processing on each fragmented data set to generate fragmented grids, and carrying out multi-beam sounding data quality real-time evaluation. According to the method, the gridding processing of mass and large-area discrete point data of the multi-beam sounding data can be quickly and accurately realized, the problems of real-time quality evaluation of the mass multi-beam sounding data and the like are solved, and the method can be widely applied to the field of multi-beam sounding data processing.

The invention discloses a multi-process machining process high-temperature intelligent prediction method based on sparse sensing extension, and relates to the technical field of intelligent manufacturing, and the method comprises the steps: constructing a three-dimensional simulation model of a key part of a machine tool, carrying out the finite element thermal analysis, screening nodes in an initial layout network based on a graph node centrality measurement algorithm, and laying sensors, forming a sparse sensing network; acquiring a sparse temperature sensing data set based on the network, and setting a global temperature prediction threshold and an auxiliary anomaly judgment threshold in combination with material characteristics, finite element results and historical data; machine tool numerical control system parameters are collected and analyzed in real time, a process knowledge graph is constructed, and a current process type is judged through a model; inputting the sparse data set and the process type into a space-time extension model to obtain temperature data of all positions; and the process type and the overall data set are input into a high-temperature intelligent prediction model, and the future high-temperature abnormal state is judged in combination with a threshold value, so that the high calculation cost of a mechanism driving method is avoided, and the real-time monitoring and online decision-making requirements are met.

The invention relates to a FgFisNET agricultural product recommendation method for fine-grained feature interactive selection, which is used for fine-grained feature interaction and important feature interaction selection. According to the method, firstly, effective feature interaction is learned by combining an inner product and a hadamard product, and effective modeling is carried out on feature interaction in a sparse data set; then introducing a group of architecture parameters for each group of feature interaction, learning the relative importance of each feature through gradient descent, deleting unimportant feature interaction, reserving the architecture parameters as attention units to help learning of the model, and finally performing CTR prediction of agricultural product e-commerce by using the trained model. According to the method, fine-grained feature interaction and important feature interaction selection are fused, and the performance of the model is improved. The effectiveness of the method is verified through a real agricultural product e-commerce platform, and remarkable economic value is achieved.

This disclosure provides a data-driven and scalable method to discover cause-and-effect relationships in data from natural systems that include sparse data sets. This technique can learn a causal graph from heterogenous data sources by combining embeddings from real data and embeddings from simulated data generated by process-based models. The causal graph is used for what-if analysis in out-of-distribution settings. One application is understanding the factors that affect soil carbon. A causal model created by these techniques can be used to discover cause-and-effect relationships that affect soil carbon. This model has applications such as forecasting soil carbon for a future time point to help inform farm practices. Farm practices, like tilling, may be modified in response to predictions provided by the model.

The present invention relates to the technical field of geotechnical parameter analysis, and discloses a sparse data probability distribution modeling method applied to geotechnical parameter data acquisition. A sparse data set is established according to the collected sample data and the number of samples; based on the sparse data set, the Gaussian function is used as the kernel function, and the approximate probability density function and the approximate cumulative distribution function are obtained through the kernel density estimation algorithm; the empirical cumulative distribution function of the sparse data set is obtained; the maximum absolute difference is obtained through the difference between the approximate cumulative distribution function and the empirical cumulative distribution function; a number of random bandwidth parameters are used as the initial population of chromosomes, and the maximum absolute difference is optimized and solved through the genetic algorithm, and the chromosome corresponding to the maximum fitness is obtained as the optimal bandwidth parameter of the approximate probability density function and the approximate cumulative distribution function; the optimal bandwidth parameter is substituted into the approximate probability density function and the approximate cumulative distribution function to obtain the target approximate probability density function and the target approximate cumulative distribution function.

Techniques are described for improving driver efficiency. An example method can include a device accessing sparse location data indicative of one or more geographic locations along a route of the user device during a first time period. The route includes a starting location data point and an ending location data point. The device can access motion data collected by the sensors of the user device. The motion data can be collected by the sensors during the first time period. After a conclusion of the first time period, the device can generate, using the sparse location data and the motion data, a dense data set to reconstruct a route that includes the starting location data point and the ending location data point. The reconstructed route can include second dense location data and velocity data. The device can store the reconstructed route in a local memory of the user device.

Methods and systems for utilizing transfer learning to deal with sparse datasets in wireless networks, particularly those associated with system-level network modeling. The methods and systems are designed to overcome the limitations posed by the sparsity and dynamicity of real network data, which is often difficult to collect due to the substantial costs and potential performance degradation associated with conducting system-level experiments on large numbers of base stations.

Systems, computer-implemented methods, and computer program products for facilitating characterization of crosstalk of a quantum computing system based on a sparse data set are provided. According to one embodiment, a system can include a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can include a packing component that packs a subset of quantum gates in a quantum device into one or more bins. The computer executable components can also include an evaluation component that characterizes crosstalk of the quantum device based on a number of bins into which the subset of quantum gates is packed.

This invention provides a method for predicting the spatiotemporal evolution of air quality in urban built-up areas and for forecasting exposure risks. It relates to the field of electronic digital data processing technology. The spatiotemporal evolution method includes: discretizing the target area to be monitored into a cubic sampling domain; planning the flight path of a UAV; and constructing a sparse dataset containing five-dimensional features. Iterative upsampling operations are performed to transform the sparse dataset into a dense set of feature points covering the sampling domain. The pollutant transport flux vector of each node within the dense feature point set is calculated, and a weighted reverse tracing vector pointing to the suspected source is constructed. A three-dimensional source tracing probability cone is then constructed by combining turbulence intensity. The centroid offset coefficient and axial dispersion are extracted, and the emission condition label of the pollution source is identified using a ternary emission condition discrimination model. Multi-source probability field superposition and fusion are performed to determine the estimated coordinates of the pollution source. Based on the estimated coordinates, a forward spatiotemporal diffusion simulation is performed, and a three-dimensional risk warning map is output.

The present invention discloses an industrial defect classification method based on federated transfer learning, which belongs to the technical field of federated learning and industrial defect classification, including data preparation, a training client locally collects and preprocesses defect data generated in the industrial production process, and performs data processing; local model training, the training client inputs the data processed by step S1 through a high-low dual-channel multi-head attention mechanism module until the local model converges; model update and upload; model weight calculation and aggregation; model deployment and inference, the target client receives the global model sent by the server, and is used for industrial defect classification of the data to be tested. The present invention adopts the above-mentioned industrial defect classification method based on federated transfer learning, and exhibits strong transfer generalization performance in terms of protecting privacy and processing defect classification tasks with imbalanced small sample categories in multi-source domains, providing a new idea for defect classification in scenarios where data sets are scarce and privacy protection needs to be enhanced.

The invention discloses a liquid drop parameter extraction method based on a light scattering theory constraint neural network, and the method comprises the following steps: firstly simulating and obtaining an accurate standard rainbow signal of a liquid drop according to a Lorenz-Mie theory, and constructing a network data set; constructing a rainbow inversion network; then establishing an index function for evaluating the rainbow inversion network droplet parameter extraction performance; carrying out gradient descent by using a loss function based on a light scattering theory to update parameters of the rainbow inversion network, and selecting an optimal rainbow inversion network by using a verification set and an evaluation function; and finally, inputting the test set into the rainbow inversion network to obtain the particle size and refractive index parameters of the liquid drops, and testing the liquid drop parameter extraction performance of the rainbow inversion network. According to the method, the droplet parameter extraction capability of the rainbow inversion network under the condition of sparse data sets is enhanced, and the method can be applied to the fields of fuel injection, spray cooling, drying and the like related to droplet multi-parameter measurement.

The invention belongs to the technical field of traffic generation, and discloses a diversity-oriented industrial protocol format inference traffic generation method. Designing an adversarial flow generation model oriented to format inference, and isolating learning of flexible payload distribution from rigid syntactic enforcement by adopting a decoupled generation architecture; the generation process is separated from a deterministic rule, so that the traffic generation model can synthesize a high-entropy payload mode exceeding the finite diversity of an original sparse data set; by isolating flexible distributed learning from rigid syntactic enforcement, the FIGAN solves the internal conflict between syntactic effectiveness and semantic diversity, and a new normal form is provided for industrial traffic enhancement.

The methods discussed herein can extract relevant signals from sparse data sets, for instance in cryptographic analysis, noise reduction, pattern recognition, or computational genetics. The present solution can improve technological performance of an analytical device such as through reducing server load, computation time, and data storage sizes. The present solution can identify relevant signals, such as genetic variants with a high probability of pathogenicity, in large, sparse data sets.

This disclosure provides a mechanism for the enrichment of sparse datasets using language models. By training language models on the specific distribution of known values in a dataset, missing values can be predicted, and the predicted values added, thereby resulting in a more complete dataset. This method also facilitates the enhancement and augmentation of datasets by predicting values for new properties that were not previously available. The approach proves particularly effective at scale, transforming large sparse datasets into more complete and enhanced datasets. Masking language modeling may be employed to train language models capable of generating representations of technical data. Training data includes corpuses of technical data that may be represented as text strings. These pretrained models are fine-tuned to predict various properties. The resulting models can predict missing values in large technical datasets, providing valuable data for guiding scientific research.

A system may access a set of training data and determine a timeframe associated with a positively labeled data item of the training data. A system may generate at least two new positively labeled data items based on the positively labeled data item to generate augmented training data. A system may train a machine learning model by applying the augmented training data as input to a machine learning model, and modifying a weight of the machine learning model.

The present invention discloses a high-dimensional sparse image representation method based on a deep semi-supervised learning framework, comprising the following steps: 1) initializing the Semi-DRRBM model parameters θ; 2) obtaining the sample hidden layer features; 3) using the reconstructed visible layer data; 4) using the obtained sample reconstructed hidden layer features; 5) calculating the partial derivatives of H(θ; G) and H'(θ; G'); 6) obtaining the current calculation with the previous hidden layer data h and visible layer data v, the reconstructed hidden layer data h' and visible layer data v', and the parameters obtained in the i-th calculation; 7) when new parameters θ are obtained through step 6, taking them as new inputs and performing the iteration of steps 2-6 again; 8) returning the parameters θ of the Semi-DRRBM model and substituting them into the Semi-DRRBMRBM model to complete dimensionality reduction. The present invention can extract low-dimensional and dense hidden features on high-dimensional and sparse data sets.

This disclosure provides a mechanism for the enrichment of sparse datasets using language models. By training language models on the specific distribution of known values in a dataset, missing values can be predicted, and the predicted values added, thereby resulting in a more complete dataset. This method also facilitates the enhancement and augmentation of datasets by predicting values for new properties that were not previously available. The approach proves particularly effective at scale, transforming large sparse datasets into more complete and enhanced datasets. Masking language modeling may be employed to train language models capable of generating representations of technical data. Training data includes corpuses of technical data that may be represented as text strings. These pretrained models are fine-tuned to predict various properties. The resulting models can predict missing values in large technical datasets, providing valuable data for guiding scientific research.

The invention belongs to the technical field of data processing, and particularly relates to a heuristic data classification algorithm based on a random walk strategy, which comprises four steps of preprocessing original data and initializing algorithm parameters, generating a similarity chain based on a random walk algorithm, segmenting the similarity chain and determining a label of the similarity chain, and determining a label of a data point. According to the method, a brand new thought for realizing data classification is provided, the defects of the prior art are overcome, the design of the technical framework is simple and efficient, the implementation is easy, and particularly, the defect that the existing classification algorithm is poor in sparse data set classification effect is made up for. Benefited from the two-stage independent design, the algorithm is easy to improve and secondarily develop, and has better generalization ability and expandability. Experimental cases show that the method has good practical utility in classification of various data, enriches original algorithms in the field of data classification, and provides new reference and support for data processing.

An ensemble time series prediction system that makes predictions based on observed data. The disclosed ensemble time series prediction system may leverage different types of datasets and information from different resources for making predictions. The disclosed ensemble time series prediction system may extract time dependent features from autoregressive time dependent data, embedding features from sparse datasets, continuous features from continuous dataset, and time lagged features from data that include time-lag information. The disclosed ensemble time series prediction system may then consolidate the features extracted from the different types of datasets and generate a set of consolidated input features for training a neural network, which may include a recurrent neural unit that finds sequential pattern for the sequence of input features and a regression unit that performs regression and predictions. The ensemble time series prediction system may output a set of outputs that include predicted values and associated confidence intervals.

The methods discussed herein can extract relevant signals from sparse data sets, for instance in cryptographic analysis, noise reduction, pattern recognition, or computational genetics. The present solution can improve technological performance of an analytical device such as through reducing server load, computation time, and data storage sizes. The present solution can identify relevant signals, such as genetic variants with a high probability of pathogenicity, in large, sparse data sets.

The invention discloses a tooth profile parameter optimization design method based on Gaussian regression. The method comprises the following steps of: 1, performing gear finite element meshing contact simulation setting, including establishing a gear pair model, establishing a contact pair, performing grid division and setting boundary conditions; 2, solving and monitoring a force convergence curve in real time through gear pump meshing contact simulation setting in the step 1, and further obtaining and analyzing the contact stress change of gear dynamic meshing; step 3, constructing a Gaussian regression model by taking the sparse data set as a training set through the gear meshing simulation data set in the step 2, predicting maximum contact stress data under all tooth profile parameter combinations, further obtaining an optimal parameter combination, and carrying out accuracy verification; according to the method, the calculation cost can be remarkably reduced, and the gear tooth profile parameter optimization design efficiency is greatly improved.

A method may include obtaining three-dimensional coordinate points and a two-dimensional image representing an environment and objects in the environment. The method may include segmenting the image into image segments and obtaining a pixel selection in the image. The method may include generating an image segment frustum corresponding to a particular image segment shape that includes the pixel. The method may include intersecting the image segment frustum with the coordinate points to determine a subset of the three-dimensional coordinates used to fit a derived surface. The derived surface may represent a three-dimensional volume having a surface shape corresponding to the shape of the image segment frustum and may include a subset of three-dimensional coordinate points that intersects with the image segment frustum. The method may include generating a derived three-dimensional coordinate point that represents a surface point included in the environment within a volume of the derived surface.

The application discloses a kind of unmanned aerial vehicle hyperspectral non-imaging paradigm's oil slick detection method, comprising: parallel push-scan type scanning is carried out to offshore oil spill area, and the hyperspectral original data of measurement sea area is obtained;Spectral mixing matrix pixel splicing method is used to seamlessly splice the hyperspectral original data into image to obtain clean sea area data, other coastline and land non-oil slick area hyperspectral pixel cube;Full-spectral data set obtained in oil slick area and suspected oil slick area is regarded as foreground, and corresponding area relative position information is retained, other clean sea area data, other coastline and land non-oil slick area data are marked as background;The background is carried out hyperspectral cube sparse representation, and spectral segment sparse line compression is carried out using matrix mapping mode to obtain sparse data set, and the feature data set of oil slick area and sparse data set are used as transmission data;Containing relative position information oil slick area is marked to background image, and oil slick area focusing result is marked on map.

The invention provides a regression training method and device based on a scale-inconsistent noisy sparse data set, relates to the field of animal feeding, and solves the problem that the prior art cannot work well on a scale-inconsistent noisy sparse regression label data set. And the technical problem that the deep learning network cannot converge or the final performance is poor in the training process is solved. The method comprises the following steps: acquiring a training data set D, and dividing the training data set D to obtain sub-data sets; performing dual-path joint training on the sub-data set through a preset network model to obtain a calibration regression value and a prediction regression value of each animal image sample; and performing iterative training on a currently trained preset network model based on the first loss function and the second loss function to obtain a trained image regression network model. The method is used in the animal body condition scoring process.

The present invention relates to a method for locating and identifying personal identification information based on static contamination, and belongs to the field of network information security technology. This method uses input personal identification information as shared information, and uses the behavior of shared information flowing in the data to discover the association between different "service-locations", construct an information flow graph, and use the traversal algorithm of graph theory to locate and extract personal identification information. This method automatically and accurately locates and extracts personal identification information based on the input infection value. The present invention effectively solves the problem of accurately locating and extracting personal identification information in network traffic data, especially large-scale network traffic data, while overcoming the false positives caused by "excessive contamination", and can help automatically label large-scale sparse data sets, avoiding manual intervention and crowdsourcing feedback.