21 results about "High dimensionality" patented technology

A power load prediction method, system and device based on GHOA-CNN-iTransform-LSTM, and the method comprises the steps: firstly obtaining the multi-source data of a power load and analyzing the causal relationship, then constructing an analysis model based on scene injection, constructing an improved hiking optimization algorithm GHOA based on a hierarchical strategy evolution framework, optimizing the hyper-parameters of the analysis model, and finally obtaining the multi-source data of the power load. And finally, after the analysis model is trained, power load prediction of the smart power grid is realized. By acquiring the multi-source data, analyzing the causal relationship and adaptively transforming the kernel of the analysis model, the analysis model can integrate the time and the causal logic for dynamic allocation, accurate perception of external events is realized, a stable deduction result is obtained from complex data, and the accuracy of the deduction result is improved. According to the method, the self-adaptive optimization of the optimal parameters is realized through the hierarchical strategy evolution framework, the search strategy is ensured to dynamically adapt to the high dimension and complexity of the parameter space, and the precision and adaptability of power load prediction are ensured.

Systems and methods are provided for reducing a number of performance metrics generated by network functions to a number of reduced dimension metrics, which can be used to detect anomalous behavior and generate a warning signal of the detected anomalous behavior. The disclosed systems and methods transform raw performance metrics in a high dimensionality space to a reduced number of metrics in a lower dimensionality space through dimensionality reduction techniques. Anomalous behavior in network performance is detected in the high dimensionality space using the reduced dimension metrics. The systems and methods disclosed herein convert the reduced dimension metrics back to the high dimensionality space, such that the performance metrics from network functions can be utilized to understand and address potential problems in the network.

The present disclosure relates to methods, systems, and apparatuses for featurizing time-series data to enhance machine learning model training. Time-series data, such as transaction records, is preprocessed to identify fields including descriptive and categorical information. Categories are assigned using a first machine learning model, and tags are applied based on domain-specific patterns or large language models. The processed data is organized into a star schema data structure comprising a fact structure and associated dimension structures. Features are generated from the data structure based on time windows, incorporating statistical metrics and identified patterns. These features are provided to a machine learning module to train a second machine learning model, improving accuracy and adaptability for applications such as customer behavior prediction and financial analysis. The disclosed approach addresses challenges of high dimensionality, noise, and temporal dependencies in time-series data, enabling robust and contextually relevant feature generation.

The application discloses a hyperspectral image classification method and system based on a quaternion deep network, mainly strengthens the learning ability of the network and improves the training efficiency of the network through quaternion algebra theory and deep learning theory, and solves the problems of few hyperspectral image data samples, spectral space variability and the like.The implementation scheme is as follows: firstly, considering the high dimensionality of data, dimensionality reduction is carried out through principal component analysis.Next, a deep attention module is constructed, important features are strengthened and noise interference is suppressed, so that the quality of data is improved.Then, a multi-branch network is constructed, the strengthened features are grouped, and a quaternion generator is designed to map the hyperspectral image data to a quaternion space for processing.Meanwhile, a multi-scale quaternion attention module is embedded to further strengthen the feature expression capability.Finally, the features of each branch are fused, and high-precision classification of the hyperspectral image is realized.

The invention discloses a transaction fraud detection method, system and device and a storage medium, belongs to the field of financial fraud detection, and solves the technical problem that in the prior art, when transaction data with high dimensionality, small samples and unbalanced extreme categories are adopted, only shallow combination of homogeneous integration and simple migration is adopted, and performance improvement in a cross-domain scene is limited. Obtaining a public credit card fraud data set, pre-training the deep neural network structure to obtain a feature extractor, and extracting migration features; splicing the migration features and fraudulent transaction detection data features to construct a combined data set; designing a base model layer composed of a plurality of heterogeneous base models to predict the combined data set, and training the meta-model layer based on a prediction result to obtain a TEL integrated model; and performing fraud detection on the transaction data by using the TEL integration model to obtain a fraud detection result. The method is used for realizing high-accuracy and high-stability transaction fraud detection.

Mechanisms for training a distributed neural network are provided, the mechanisms including: for each of a plurality of sub-networks: performing, using a hardware processor, a transform on data in a data structure having at least two dimensions to provide training data having a higher dimensionality than the at least two dimensions; and training the sub-network using the training data independently of other of the plurality of subnetworks. In some of these embodiments, the at least two dimensions is two dimensions. In some of these embodiments, the training data is stored in a three-dimensional structure. In some of these embodiments, the transform is a discrete Fourier transform. In some of these embodiments, the transform is a discrete cosine transform.

A method and system(s) for processing queries to sparsely populated datasets. The method includes (i) receiving a query from a client system, (ii) determining, based at least in part on the query, a plurality of partitions of a hypercube that are expected to comprise information used in generating a response for the query, (iii) communicating a set of requests to a data source storing the hypercube, the set of requests comprising one or more requests corresponding to the plurality of partitions of the hypercube, (iv) obtaining, from the data source, a set of results to the set of requests, the set of results comprising data stored in the plurality of partitions of the hypercube, (v) determining a plurality of virtual objects for the set of results, (vi) combining the plurality of virtual objects into a response to the query, and (vii) providing the response to the query.

This invention discloses an early fault diagnosis method for high-speed train traction transformers with missing data, comprising: (1) optimizing the individual position iteration equation and initial population position of the slime mold algorithm using Lévy flight and optimal point set distribution, and optimizing the running structure of the slime mold algorithm using a teaching optimization algorithm; (2) combining the improved slime mold algorithm with a CatBoost classifier to improve the F1 score of multi-class classification and finding the optimal CatBoost parameter configuration for classification performance; (3) training and classifying transformer fault data using a model combining the slime mold algorithm and the CatBoost classifier. This invention overcomes the influence of the large scale, high dimensionality and nonlinearity of transformer data on multi-class classification and the imbalance of the number of each category. Addressing the problem of mutual influence between different CatBoost parameter settings, it uses optimization algorithms to find the optimal parameter settings for classification performance. Finally, starting from the structural characteristics of the slime mold algorithm, it improves the problem of slow convergence speed and easy getting trapped in local optima in the traditional slime mold algorithm.

The application relates to the technical field of computer vision and image processing, and discloses an image interpolation method based on high dimensionality, which comprises a video decoder module, a timestamp analysis module, a feature extraction module, a time sequence correlation modeling module, a time sequence correlation modeling module, a convolutional neural network module, an edge smoothing processing module, an edge detection module, an edge enhancement module, an image output module and a quality evaluation module; the video decoder module extracts images; the timestamp analysis module records time points and positions; the feature extraction module extracts image edge feature data; the time sequence correlation modeling module captures time and space correlation data; the convolutional neural network module performs pixel-level interpolation on image frames; the edge smoothing processing module optimizes the interpolation results; the edge detection module identifies pixel points; the edge enhancement module is applied to perform enhancement processing on edge pixel points; the image output module generates high-quality interpolated image frames; and the quality evaluation module checks the quality of the generated images.

This application discloses a grounding grid corrosion diagnosis method, device, equipment, and medium based on sensitivity analysis and dimensionality reduction, relating to the field of power equipment fault diagnosis technology. The method equates the grounding grid to a resistive network. With the goal of minimizing a nonlinear diagnostic model, an objective function is constructed based on the initial resistance amplification factor of each branch and the sum of squared residuals between the measured and calculated port voltage values. An improved Hippo optimization algorithm is used to minimize the objective function, obtaining the resistance amplification factor of each branch. The sensitivity of each branch resistance to the objective function is calculated, and it is determined whether the sensitivity value is less than a preset threshold. If it is less, the branch is marked as a healthy branch. The improved Hippo optimization algorithm is used iteratively to solve for the other branches except for the healthy branches until the convergence condition is met, obtaining the target resistance amplification factor. The grounding grid corrosion state is diagnosed based on the target resistance amplification factor. This method is suitable for grounding grid systems with high dimensionality and sparse corrosion branches, accurately locating corrosion branches and their degree of corrosion.

A method for enhancing detection of morphed biometric modality data is provided that includes the steps of receiving, by an electronic device, biometric modality data of a person and extracting, using a trained machine learning model operated by the electronic device, high dimensionality feature vectors from the received biometric modality data. Moreover, the method includes the steps of generating a distribution from the high-dimensionality feature vectors, calculating a deviation between the generated distribution and a corresponding record high-dimensionality feature vector distribution of the person, and comparing the calculated deviation against a threshold deviation value. In response to determining the deviation satisfies the threshold deviation value, the method includes the step of determining the received biometric modality data was morphed.

The invention discloses a voltage prediction method and system suitable for a power distribution network under distributed aggregation resource access, and the method comprises the steps: carrying out the partitioning of the power distribution network based on the dynamic response of a distributed power supply; classifying multi-source data, including time series data and multi-dimensional heterogeneous data, for the divided regions; applying an improved CSO-OIF Elman model to the time series data to predict the voltage fluctuation in the region in real time; and a residual error threshold filtering and switching mechanism is established, and when the prediction residual error of the improved CSO-OIF Elman model exceeds the standard, the voltage fluctuation prediction result of the XGBoost model is driven based on the multi-dimensional heterogeneous data. According to the method, the complementarity of the multi-source data is fully utilized, the time sequence dynamic state of the voltage under DG access is captured, the nonlinear correlation of the multi-dimensional characteristics is mined, compared with a single model, the voltage prediction precision can be remarkably improved, and the method is particularly suitable for a power grid scene with large distributed power supply output fluctuation and high multi-source data dimensionality.

The application provides a proxy-assisted collaborative particle swarm optimization algorithm for expensive large-scale constrained optimization problems, and an urgent need for efficient processing of expensive large-scale constrained optimization problems in the field of intelligent manufacturing and the like. The existing methods often perform poorly when facing high dimensionality, high calculation cost and complex constraints at the same time. The application decomposes a complex optimization problem into a main-aid problem for collaborative solution by splitting large-scale decision variables, so as to overcome the curse of dimensionality. A hybrid proxy model management mechanism of the main-aid problem is designed, the model precision and the calculation resource allocation are dynamically optimized, and the calculation cost is reduced. A multi-population collaborative evolution optimization algorithm of the main-aid problem is given, and the solving efficiency is significantly improved by introducing an information migration and feedback mechanism. The method is applied to X benchmark test problems and an aero compressor blade aerodynamic optimization design problem, and is compared with x kinds of popular algorithms, and the effectiveness of the method is verified by experiments.

The application provides a hyperspectral band selection method and system based on hypergraph learning, and belongs to the field of image information processing. In order to solve the problems that the existing high-dimensional data calculation is complex, and the redundant information existing between adjacent bands may cause information overlap and noise interference in subsequent problem analysis. The application comprises self-representation learning on hyperspectral data; then hypergraph learning is carried out, the hyperspectral image is segmented, hypergraph definition is carried out, and the node-node relationship in the hypergraph is described by using a hypergraph Laplace matrix; a unified framework of the joint self-representation model and hypergraph learning is designed, and the hyperparameters are optimized through a hyperopt library; finally, the update of the selection matrix P is carried out, and the hyperspectral image bands are selected according to the update result. The application considers the high dimensionality of the hyperspectral image, the insufficient training sample and the data structure and matrix information entropy problem, has the characteristics of good reliability and relatively high precision, and is suitable for popularization and use.

A system, method and computer program product for analyzing data of high dimensionality (e.g., sequence reads of nucleic acid samples in connection with a disease condition) are provided.

This invention, entitled "A Visual Guidance Robot Unordered Grasping Planning Method Based on Deep Reinforcement Learning," belongs to the field of robot grasping planning technology. Addressing the problems of existing methods relying on local optimum scoring in complex scenarios, which easily leads to frequent misgrasping, and the high dimensionality of end-to-end learning exploration, this invention proposes: acquiring sensor data to extract foreground point clouds, generating a set of grasping candidates filtered by physical feasibility; constructing reinforcement learning state vectors and inputting them into a policy network, with discrete branches outputting target candidate indices and continuous branches outputting pose and execution parameter residuals; obtaining refined pose parameters after action calculation, and then executing the grasping operation. This invention compresses the decision space to the candidate neighborhood, balancing exploration efficiency and long-term continuous box clearing benefits, and is mainly used for grasping unordered scattered objects in industrial warehousing and logistics sorting.

The invention discloses a remote sensing image classification and target detection method fusing three-way decision and multi-agent reinforcement learning, and belongs to the technical field of remote sensing image processing. The method comprises a classification process and a target detection process: in the classification process, firstly, feature importance is calculated through grey correlation analysis, then a multi-agent system is constructed to optimize three decision threshold values of a TwGrey feature selection algorithm, and after an optimal feature subset is obtained, the optimal feature subset is input into an SVM classifier to complete classification; according to the target detection process, firstly, a multi-agent collaborative improved FPN is constructed to strengthen multi-scale features, then, a high-quality anchor frame is screened through a sequential three-way decision model, and finally, a detection head is input to complete target detection. The system is correspondingly provided with a classification module, a target detection module, a multi-agent reinforcement learning module and a data interaction module, and cooperative work among the modules is achieved. The method effectively solves the problems of high remote sensing image classification dimension, multi-scale target detection and sample imbalance, improves the processing precision and stability, and can be widely applied to the fields of urban planning, geological disaster monitoring and the like.

The invention relates to the technical field of intelligent detection, in particular to a natural gas pipeline leakage detection method and related products, and the method comprises the steps: constructing an observation state data matrix; constructing an observation state data submatrix; constructing a convolutional long-short-term memory self-encoding model; an optimal convolution long and short term memory self-encoding model is obtained; the final pipeline operation characteristics are obtained; obtaining k feature clusters representing operation distribution characteristics of different pipelines; k OCSVM models are constructed; constructing a real-time observation state data matrix; k decision function values of leakage anomaly detection are obtained, and integrated decision making is carried out in a probabilistic mode; in order to solve the problem that complex characteristics such as nonlinearity, high dimension, multi-distribution and time dependence coexist in pipeline SCADA operation data, so that a traditional data-driven leakage detection method is poor in performance, a convolutional neural network and a long-short-term memory network are adopted to construct a convolutional long-short-term memory self-encoder. The method can effectively obtain long and short term dependence features of pipeline multivariable time series data in a low-dimensional feature space, improves the learning capability of the model for important features of process data, learns the multi-distribution features of pipeline operation data by using feature clustering on this basis, overcomes the leakage detection limitation of a single data distribution hypothesis model, and improves the detection efficiency. Therefore, the pipeline operation risk monitoring and control process is safer and more reliable.

Computer-implemented method for digital interaction optimization that includes: receiving an input signal for a present digital interaction; retrieving stored value functions corresponding to previous digital interactions having a matching use case to the present digital interaction, each of the stored value functions including stored data values for a plurality of parameters; generating a present value function based at least in part on present data values for the plurality of parameters, the present data values being extracted from the input signal; evaluating the present value function against each of the stored value functions to generate corresponding outputs; determining, based on at least one of the outputs, a preferred value for one or more of the parameters; and executing a subsequent digital interaction having the matching use case based on the preferred value.

This invention discloses a multi-agent collaborative incentive method based on large-model-assisted dynamic grouping, relating to the fields of multi-agent collaboration and artificial intelligence. This invention addresses problems in multi-agent collaboration under complex environments, such as insufficient team cohesion, difficulty in spontaneous alignment of individual goals, and excessively high dimensionality of the joint policy space caused by large-scale agents, leading to low training efficiency and convergence difficulties. It introduces a large language model to achieve global situational awareness and dynamic grouping, decoupling complex global collaboration into local small-group collaboration and global coordination. A follow-consistency reward function is designed to guide member agent actions to align with the leader's behavior. Combined with task rewards, a comprehensive reward is constructed, simplifying collaborative learning into a hierarchical model of leader exploration and member following, improving training efficiency and team collaboration stability.

This application provides a method, apparatus, electronic device, and storage medium for generating a planned trajectory. The method includes: acquiring a state space and the vehicle's action space; classifying the predicted trajectories of neighboring vehicles based on potential collision risks and the action space to obtain a set of risk trajectories and a set of safe trajectories; and modeling the state space, action space, set of risk trajectories, and set of safe trajectories using a partially observable Markov decision process to obtain a reference trajectory. The decision-making and planning process under the uncertainty of neighboring vehicle behavior is modeled as a POMDP, thereby fully predicting the uncertain behavior of neighboring vehicles and ensuring the driving safety of the vehicle. Furthermore, considering the excessively high dimensionality of the predicted trajectory clusters of multiple traffic participants in complex scenarios, the POMDP model is prone to the curse of dimensionality. Therefore, by classifying the predicted trajectories of neighboring vehicles into a set of risk trajectories and a set of safe trajectories based on potential collision risks, the dimensionality of the state space is reduced, effectively improving the efficiency and stability of decision-making and planning.