675 results about "Data selection" patented technology

The invention relates to the technical field of language model integration, in particular to a large language model integration method supporting semantic correction. The method comprises the following steps: S1, selecting several large language models, and sorting and preprocessing text data for training and testing; s2, training each source model and an optimization model by using an intelligent fusion language model technology; s3, fusing parameters of each source model by using an intelligent fusion language model technology, selecting model parameters for fusion according to the text data, and adjusting the fused model; s4, evaluating the semantic correction capability of the fused model, and adjusting a fusion strategy and regularization parameters according to an evaluation result; and S5, integrating the trained model into a target system. According to the large language model integration method supporting semantic correction, through an intelligent fusion language model technology, a context sensing and hybrid regularization technology is utilized to train and optimize a model and enhance adaptability, and a FuseLLM fusion technology is adopted to merge all source model parameters.

The invention discloses a personalized acupuncture treatment scheme recommendation method and system based on data analysis, and relates to the technical field of acupuncture. The method comprises the following steps: constructing a feature vector matrix according to multi-dimensional feature data of a target patient; taking the matrix as a constraint, retrieving similar historical cases in a pre-constructed acupuncture case database, and obtaining a sample case set and a similarity set; configuring a data selection strategy based on the similarity set, constructing a sample training set, and training to generate an acupuncture parameter evaluation plug-in; and defining an acupuncture parameter optimization space based on the sample case set, performing iterative optimization in the space by using the evaluation plug-in, and finally outputting an acupuncture treatment parameter scheme adaptive to the target patient. According to the method, the whole-process personalized recommendation of the acupuncture treatment scheme from case screening, model evaluation to parameter optimization is realized, and the accuracy, objectivity and clinical operability of treatment are improved.

The embodiment of the invention provides a snow depth inversion method and system based on a random forest. The snow depth detection method is applied to the technical field of snow depth detection, and comprises the following steps: selecting an acquisition area, acquiring snow depth data in a predetermined acquisition area, and preprocessing to obtain processed data; selecting a prediction variable and a target variable, training the random forest model to obtain a trained random forest model, and performing precision verification; and inputting snow depth data acquired in real time into the trained random forest model, performing snow depth prediction, and outputting to obtain a prediction result. In this way, active microwave high-resolution snow depth data and passive microwave inversion of the mountain snow depth are combined, and Sentinel-1 inversion snow depth data are taken as a training target, so that the inversion precision of the mountain snow depth is improved.

The invention relates to the technical field of data backup, and discloses an iOS device data backup and recovery method and system, and the method comprises the steps: carrying out the SSL certificate verification and EscrowBag key authentication of an iOS device, and obtaining a data path list; according to the data path list, parallel scanning of APFS file system B-Tree traversal, NAND flash memory page reading and SQLite WAL log analysis is carried out on the iOS device storage partition, and an original data set is obtained; performing integrity evaluation based on the original data set to obtain an effective data set and constructing a virtual file index; the data selection list of the user is confirmed based on the virtual file index, safety recovery is executed, and a data recovery result is returned. The SQLite page Cell reconstruction and fragment file signature matching technology is combined, intelligent repair and quality prediction can be conducted on damaged data, the user experience and operation efficiency are improved, and the user experience is improved. And the security and the reliability in the data recovery process are ensured.

A system may include a robotic control circuit configured to control one or more robotic functions of a robot. A system may include a plurality of sensors configured to collect data. A system may include a governance analysis circuit configured to analyze the data and select one or more governance frameworks based on the analyzed data. A system may include a governance model circuit configured to generate a model that applies the one or more governance frameworks to determine one or more governance actions, wherein the robotic control circuit is configured to control the one or more robotic functions in accordance with the one or more governance actions, wherein the robotic control circuit, the governance analysis circuit, and the governance model circuit are integrated on a single substrate.

The invention relates to a turbine blade cooling optimization method and device based on non-Newtonian fluid. Comprising the following steps: step 1, carrying out grid division and simulation processing on a three-dimensional geometric model of the micro-channel of the turbine blade containing a shear groove to obtain a simulation grid model; 2, on the basis of the simulation grid model, determining a non-Newtonian fluid constitutive model suitable for the simulation grid model, and associating the non-Newtonian fluid constitutive model to the simulation grid model to obtain a fluid characteristic grid model; step 3, calculating index data based on the fluid characteristic grid model and the target equation; 4, adjusting the structure parameters of the shearing grooves in the simulation grid model, repeatedly executing the step 1 to the step 3 according to the adjusted structure parameters of the shearing grooves, and recording index data corresponding to the structure parameters of each group of shearing grooves to obtain a candidate data set; and 5, selecting the structure parameter of the shear groove with the comprehensive optimal resistance performance and heat transfer performance from the candidate data set according to the index data, and taking the structure parameter as an optimized target shear groove parameter.

The invention discloses a federal data selection method and device based on a proxy verification data set, and the method comprises the steps: a central server selects participated clients according to the selection probability of the clients, and transmits a prediction model and meta-model parameters to the clients; the client calculates a sample and a client weight by using the meta-model, selects a training sample to update a prediction model according to the sample weight, and uploads the client weight and the updated prediction model to the central server; the client with the highest weight updates the meta-model by using the proxy verification data set, and uploads the updated meta-model to the central server; the client updates the proxy verification data set by using a meta-marginal function; and after receiving all the prediction model parameters, the central server carries out parameter aggregation to obtain an updated prediction model, and updates the client selection probability to carry out the next round of model training. According to the method, the dependence on an extra unbiased verification set is eliminated, and the calculation efficiency is high.

The invention relates to power load prediction, in particular to a dynamic hierarchical granularity model power load prediction method considering time sequence factors. According to the method, the fitting capability of periodic features, trend changes and sudden influences in load fluctuation is remarkably enhanced, and the problem of feature loss caused by neglect of a time sequence dynamic weight of an existing model can be solved. Comprising the following steps: S1, performing time sequence characteristic analysis on historical power load data, and converting an original non-stationary time sequence into a stationary time sequence through difference and logarithm transformation; s2, obtaining a cleaned power load data set; s3, based on the data cleaned in S2, selecting a Gaussian radial basis function (RBF) as a kernel function, and analyzing temperature, holiday identification and linear change trend key influence factors at the same time; s4, performing dynamic multi-level granulation on the data set, and dynamically adjusting the particle level according to the data mixing degree and the particle density; and S5, fusing the time sequence kernel function and the influence factor through a decision function, and predicting the future power load.

Systems, methods, and computer-readable storage media for measuring spinal canal volume in vertebrates, and more specifically to using Artificial Intelligence (AI) to predict how surgical options will affect spinal canal volume. A system configured as disclosed herein can receive two or more pre-operation medical images capturing at least one functional spinal unit, then calculate an initial spinal canal regional volume using the pre-operation medical images for at least a portion of the at least one functional spinal unit. The system can then calculate, using a neural network, a predicted spinal canal regional volume for at least a portion of the at least one functional spinal unit undergoing various spine surgery options separately, resulting in predicted spinal canal regional volumes corresponding to the plurality of spine surgery options. The system or a surgeon can then select, using that data, one or more of the spine surgery options.

The invention discloses a PID parameter optimization system based on offline system identification, and the system comprises a data selection module which is used for screening modeling data in a historical database; the data preprocessing module is used for performing real data complementation and data filtering on the screened modeling data; the model system identification module is used for establishing a controlled object model by using the preprocessed data; and the PID parameter optimization module is used for performing off-line optimization of PID parameters on the controlled object model. The defects in the prior art can be overcome, any on-line closed-loop test is not needed, and the optimal PID parameter is output at a time.

The invention provides a standardized instant response system and method for a resource scheduling high-concurrency scene. The system comprises an associated service verification module which is used for monitoring the change condition of service data in real time, selecting a verification chain of the highest level corresponding to the service data to execute verification for the changed service data, determining the state information of a scheduling voucher associated with the verification chain according to the verification result, and sending the state information to the scheduling voucher; the state information of the scheduling voucher is provided for the scheduling verification and response module; and the scheduling verification and response module is used for returning standardized response data to the user side according to the state information of the target scheduling voucher provided by the associated service verification module in advance after the target scheduling voucher in the resource scheduling request is obtained through analysis. Through the technical scheme provided by the invention, the associated service verification time consumption can be greatly reduced, the resource scheduling full-chain duration is reduced, and the resource scheduling efficiency is greatly improved.

The invention discloses a question and answer intention recognition and classification method for an infant care robot. A question rewriting node, an intention recognition node and a branch node are included. The input data of the question rewriting node are questions input by a user, the training data of the question rewriting node are phrases for mothers, infants and child rearing scenes, the model parameters of the question rewriting node are obtained through adjustment based on the training data, and the question rewriting node processes the input data according to the model parameters of the question rewriting node and outputs the data; the intention recognition node obtains data output by the problem rewriting node and recognizes a user intention and an expected result; the branch nodes provide a plurality of large language models in mother and child breeding scenes, and the branch nodes select the corresponding large language models to execute data processing according to output data of the intention recognition nodes. Original user input is clarified through standardization and rewriting, then the accuracy and response efficiency of subsequent intention recognition nodes are remarkably improved, and finally the overall user experience is improved.

Methods, systems, and apparatus for training a smaller machine learning model through contrastive learning. The method includes obtaining data specifying a larger machine learning model, wherein the larger machine learning model has been trained through contrastive learning; obtaining a training dataset comprising a plurality of training examples; and training the smaller machine learning model on the training dataset, the training comprising, at each of a plurality of training iterations: generating a batch for the training iteration that comprises a subset of the plurality of training examples, the generating comprising selecting the subset of training examples according to performing an active data selection procedure based on respective contrastive losses of the larger machine learning model on one or more candidate batches that each include a respective subset of training examples from the training dataset; and training the smaller machine learning model on a contrastive loss function using the batch.

This disclosure describes methods for adaptive machine learning in distributed edge computing. An edge node collects local data, selects a suitable large language model (LLM) or small learning model (SLM), trains it, and shares updates with a federated server or peer nodes. Another method matches AI functions with appropriate models, uses datasets with confidence values, and applies a Kalman Filter to assign weights and update covariance matrix confidence. In collaborative training, edge nodes store trained models with per-layer covariance values, transmit them to a control node, and update models based on aggregated inputs. Blockchain may be used for secure model storage and distribution, with smart contracts managing access and updates. These approaches support efficient, privacy-preserving learning by adapting models using statistical confidence and decentralized coordination.

This disclosure provides methods, devices, and systems for planning and performing medical procedures. The present implementations more specifically relate to analyzing objects in 3D images. In some aspects, a segmentation system may receive image data representing a 3D image of an anatomy, select a seed location for a target in the 3D image, and infer a segmentation mask associated with the seed location from at least a portion of the received image data based on a neural network model trained to segment a class of objects associated with the target. The system further extracts a polygon mesh from the segmentation mask to produce a 3D model of the target. The system can determine a spatial relationship between an instrument and the target based on a position of the 3D model relative to the 3D image. The system can also estimate a geometry of the target based on the 3D model.

The invention discloses a robot data analysis decision method and system based on a neural network, and relates to the technical field of robot production, and the method comprises the steps: obtaining the state data of a local robot; acquiring state data of other local robots, and selecting the robot with the highest comprehensive score as an auxiliary robot; collecting local data and auxiliary data; performing modal noise reduction, feature extraction and fusion on the local data and the auxiliary data by using the neural network preprocessing layer to obtain an environment representation vector; determining a bottom risk by using a Bayesian network; analyzing a risk evolution trend by using a knowledge graph; and inputting the underlying risk and the risk evolution trend into the reinforcement learning model, and outputting an optimal decision. According to the method, the deficiency of single-robot sensing is made up through multi-robot cooperative data acquisition, noise reduction and deep feature extraction of interfered data are realized by means of neural network preprocessing, and the environment sensing accuracy, risk assessment precision and decision reliability of the robot in the interference environment are improved.

Embodiments select features for performance prediction. In one embodiment, a method comprises: receiving a request to select features to predict a performance issue of an application, the request indicating a set of key performance indicators (KPIs) for the application and data of performance metrics; selecting a first set of features, a feature being selected to the first set of features based on correlation between the feature and the set of KPIs; selecting a second set of features from the first set of features to predict the performance issue of the application, a feature being selected to the second set of features based on a causal relationship between the feature and the set of KPIs; and causing prediction of the performance issue of the application based on the second set of features and corresponding time lags between the second set of features and the set of KPIs.

The invention discloses an ALU operation fusion processing module and method suitable for a neural network, and the module comprises a control unit which is used for receiving and decoding a machine instruction, managing the execution processes of internal and external circulation and microinstruction circulation, and generating a control signal of each stage of a microinstruction assembly line; the microinstruction buffer area is used for storing a microinstruction sequence pre-generated by the neural network compiler; the register file is used for storing source operands and results of ALU operation; each entry of the register file is composed of a valid bit, a tag bit and a data bit; the ALU computing core adopts an SIMD (Single Instruction Multiple Data) architecture and comprises a plurality of paths of parallel arithmetic logic function units; the Load / Store unit is used for processing data exchange between a register file and a local buffer area; and the data selection interface is used for selecting a data source or a target buffer area according to the storage tag field of the microinstruction. According to the method, the high efficiency and the flexibility of the ALU in the neural network hardware accelerator can be effectively considered.

The application provides a base station traffic prediction method and related equipment, and relates to the field of communication, wherein the base station traffic prediction method comprises the following steps: determining the adjacent base stations of a to-be-predicted base station according to the geographical distance information between base stations; determining the neighbor base stations in the adjacent base stations, and the historical traffic records of the neighbor base stations and the historical traffic record of the to-be-predicted base station have a causal relationship; performing feature extraction on the historical traffic records of the to-be-predicted base station and the neighbor base stations to obtain feature data; and predicting the predicted traffic of the to-be-predicted base station according to the feature data. In the screening process of the feature data object, the spatial factor is fully considered to avoid the singularity of feature data selection, and the causal relationship between the base station traffics is fully considered, so that the blindness of feature data selection is avoided, the redundant feature interference is reduced, the effectiveness of feature selection is improved, and the base station traffic prediction accuracy and efficiency are improved.

To provide a molded product removing machine capable of reliably confirming whether or not a combination of a selected mold and a removal head and selected teaching data are correct when replacing the mold. [Solution] When teaching data is created by teaching data creation unit 13, an image including at least a part of the molding mold mounted on the molding machine, an image including at least a part of the take-out head, or an image including at least a part of the molded product taken by the take-out head is automatically taken by an imager, and stored as image data corresponding to the teaching data in data storage unit 12. When performing the replacement work of exchanging molding mold M and take-out head 7, data display unit 17 displays on display 18 an image based on the image data corresponding to the teaching data selected by teaching data selection unit 13.

The invention provides an intelligent wind speed and direction prediction method and system for an unmanned aerial vehicle. The method comprises the steps of obtaining offline flight data of an unmanned aerial vehicle, performing preprocessing, obtaining offline flight optimization data, performing data primary selection and fine selection processing, obtaining an optimal flight feature set corresponding to a target wind speed component, performing model training, obtaining an offline wind speed prediction model, and transmitting the offline wind speed prediction model to a preset intelligent flight control system of the unmanned aerial vehicle. Obtaining a real-time wind speed predicted value and a real-time wind direction based on the real-time flight data of the unmanned aerial vehicle; according to the method, a three-dimensional target wind speed component differentiation modeling strategy is adopted, Pearson primary selection and RFECV fine selection are fused, and an optimal flight feature set is automatically configured for each target wind speed component, so that unmanned aerial vehicle intelligent prediction of real-time wind speed and real-time wind direction perception is realized, the system is used for realizing the method, and the system is suitable for large-scale popularization and application. The three traditional problems of large load of a physical sensor, strong model dependence and deep learning computing power bottleneck are overcome, and the precision and real-time performance of wind speed and wind direction prediction are ensured.

The invention relates to a multi-factor dynamic authentication method and device, electronic equipment and a storage medium, and the method can collect user behavior data, equipment fingerprint data, real-time threat intelligence and other multi-dimensional data in real time based on a received authentication request, carries out the evaluation of the multi-dimensional data based on a pre-trained artificial intelligence model, and improves the user experience. The method comprises the steps of obtaining a comprehensive risk score of an authentication request, determining a target multi-factor authentication strategy from a preset multi-factor authentication strategy library according to the comprehensive risk score, authenticating a user according to the target multi-factor authentication strategy, and determining authorized access or denied access to the user according to an authentication result. Due to the fact that the corresponding target multi-factor authentication strategy can be selected in combination with the multi-dimensional data collected in real time, the multi-factor authentication strength can be dynamically adjusted to complete multi-factor authentication, and user experience and safety are balanced.

A computer-implemented method for predicting cardiovascular disease risk, the method including: receiving a first patient history data comprising imaging data and / or non-imaging data for a patient at a first time point; selecting prediction report parameters defining a type of cardiovascular event and a risk prediction time scale; processing the first patient history data using a trained machine learning model configured to predict disease onset, progression, and / or regression over time, wherein the trained machine learning model is trained using patient subsets created based on patient history characteristics and outcomes; generating a risk prediction for the selected type of cardiovascular event over the selected risk prediction time scale based on the processed first patient history data; generating a risk prediction report based on the risk prediction; and outputting the risk prediction report.

The invention discloses a variant reconstruction AUV shape optimization design method based on knowledge migration. The method comprises the steps that a historical database is constructed through historical design and optimization of a single aircraft; carrying out data selection so as to construct a target domain database in a target domain task of shape design optimization of the multi-body structured aircraft; generating a target domain sampling data set through sampling; establishing an intermediate domain task and a corresponding intermediate sample set for shape design optimization of the dual-body aircraft; constructing an agent model of the target domain task and the intermediate domain task; establishing a migration integration model based on a three-domain task; carrying out iterative optimization on sample points in the target domain sampling data set by utilizing the integrated migration model, and constructing a sample pool; and constructing a multi-body structured aircraft appearance hydrodynamic simulation calculation framework, performing hydrodynamic simulation based on the sample points in the sample pool, and taking the design parameters contained in the finally obtained sample point with the minimum sailing resistance as a multi-body structured aircraft appearance design scheme.

The application discloses a transmission processing method and device, a terminal and a storage medium, and belongs to the technical field of communication. The transmission processing method of the application embodiment comprises: a terminal determining first resource information of first data, wherein the first resource information is used for re-evaluation or preemption of a target resource; wherein the first resource information is associated with first target information, and the first target information comprises at least one of the following: whether the first data is a periodic service, or whether the first data enables periodic reservation; whether periodic reservation is enabled in a resource pool; a periodicity configured by the resource pool; second resource information selected or reselected by the first data; a measurement result of the resource pool or a bandwidth part; a power saving mode of the terminal; a power level of the terminal; a discontinuous reception configuration of the terminal; third resource information configured or indicated in the resource pool; and related information of second data, wherein the second data is different from the first data.

The systems and methods disclosed herein generate responses using data retrieved by validating vector embeddings using hash signatures. An output generation request is received via a computing device and includes an input that includes a content set and a command set. The content set includes data chunks, which refers to text, audio, image, and / or video data. A subset of the data chunks that fail to associate with existing hash signatures are selected. Each selected data chunk is validated against predefined constraints that define the operative boundaries of a guideline set and subsequently assigned a unique hash signature to indicate a degree of satisfaction of the data chunk with the guidelines. Using an artificial intelligence (AI) model, a response to the output generation request is generated in accordance with the data chunks, where each data chunk is associated with a corresponding unique hash signature.

The invention discloses a browser sidebar intelligent assistant interaction method and system, and relates to the technical field of man-machine interaction, and the method comprises the steps: obtaining webpage data according to webpage context information, obtaining interaction data from a user interaction record, and obtaining user emotion data through the interaction data; generating a task recommendation queue based on the webpage data, the interaction data and the emotion data by using a multi-dimensional reasoning method, and generating candidate interaction items according to the task recommendation queue and the user emotion data; receiving an instruction input by a user through the candidate interaction item, generating an analysis result in combination with the webpage data, executing a preset task corresponding to the analysis result, and obtaining task result data; and selecting a visualization template according to the task result data and the user emotion data to output a task result, and updating the user personalized configuration and the semantic memory library. The method is used for solving the problems that traditional browser assistant task recommendation lacks pertinence, and the matching degree of interactive experience and user states and preferences is low.

A method may select, from a training data repository comprising a plurality of samples with known classifications, an initial training dataset comprising a second plurality of samples. A method may provide, as an input to a classification model, feature vectors associated with the initial training dataset and may train the classification model using the feature vectors. A method may determine a classification of each sample of a third plurality of samples using the classification model. A method may determine a difference between the determined and the known classification for each sample. A method may determine a selection weighting for each sample for based on the difference between the determined classification and the known classification. A method may select a subset from the from the third plurality of samples based on the determined selection weighting. A method may train the classification model using feature vectors associated with the subset.

A control method for an arthroplasty surgical robot includes: capturing a bone image, transmitting the image to a computer system, and acquiring image data of a bone; selecting a suitable prosthesis model and placing it on a diseased joint for matching, wherein the overlapping part between the model and the bone is the part that needs to be removed and replaced; mounting signal sources; capturing a bone image again, including the signal sources and the joint part; establishing a coordinate system using the signal sources as the origin, and calculating and memorizing an interface between the model and the bone overlapping part and the coordinates of the interface in the coordinate system; and receiving the coordinates by a surgical robot, and obtaining the positional coordinates by the surgical robot of a part of the diseased joint that needs to be removed, and performing a surgical operation by the surgical robot.