21 results about "Relevance measure" patented technology

The invention relates to the technical field of circuit board analysis, in particular to an integrated circuit defect analysis method and system, and the method comprises the steps: extracting and screening E-test data of a target product from a big data platform for integrated circuit manufacturing; performing defect classification modeling processing on the screened E-test test data; calculating a defect correlation index according to the classification accuracy and the correlation metric value of each candidate E-test item; determining a defect judgment threshold value and a numerical value distribution direction related to the defect based on a logic regression classifier result of the key test item, and generating a binary defect distribution diagram corresponding to each jointed board; calculating the image similarity between the bad jointed board graph and each good jointed board graph; judging whether the image similarity is higher than a preset threshold value or not, and adding the jointed boards of the good jointed board graph into a virtual bad jointed board list; and based on the virtual bad jointed board list, analyzing and comparing the manufacturing process flow data of the virtual bad jointed board and the original defective jointed board, and determining the potential root cause causing the defect. According to the method, the problems that in the E-test scene of the circuit board, the number of defect samples is extremely small, categories are extremely unbalanced, an experience threshold value is difficult to set, and root causes are difficult to trace can be solved.

The invention discloses a large model parameter protection method based on singular value decomposition in a trusted execution environment. Positioning a key layer of the large model through gradient sensitivity and interlayer correlation measurement in a large model fine tuning training process, and decomposing a weight matrix of the key layer by using singular value decomposition to obtain sensitive parameters and non-sensitive parameters; dividing a large model operation environment into a rich execution environment and a trusted execution environment, and generating a reversible key matrix in the trusted execution environment; when input data pass through the key layer for the first time, the rich execution environment uses non-sensitive parameter reasoning and then transmits the input data to the trusted execution environment for sensitive parameter reasoning, then the data are encrypted by using the key matrix, the encrypted data are transmitted back to the rich execution environment for reasoning, and when the input data pass through the key layer again, the rich execution environment conducts reasoning and then transmits the input data to the trusted execution environment. And the final data is output after decryption and reasoning. According to the method, the side channel attack risk is reduced, the data transmission security and the calculation efficiency are improved, and the model reasoning delay is reduced.

Techniques for modelling a questionnaire are disclosed. Textual content received in a query is converted into a set of query vectors. Relevance metric is then computed for each elementary vector, in a set of elementary vectors based on semantic similarity between each query vector in the set of query vectors and each elementary vector in the set of elementary vectors. A set of relevant elementary vectors is accordingly identified. Modelling of a questionnaire, having hierarchically-linked questions, is then triggered with set of relevant elementary vectors. The modelled questionnaire includes an opening question and a plurality of subsequent questions, each being determined based on a response associated with an immediately preceding question thereto. The subsequent questions are determined until a response to a question, from amongst the subsequent questions, provides a required insight associated with an aspect. A questionnaire delivery signal is then generated to cause rendering of the modelled questionnaire.

The application relates to the technical field of industrial Internet of Things, and discloses a multi-channel time sequence data error bound quantification method and system based on dynamic graph grouping and error distribution, which comprises the following steps: S1: acquiring multi-channel time sequence data collected at the same collection point; S2: determining a scale factor, a point-by-point maximum absolute error upper bound, normalized channel time sequence data, and normalized error bounds of all channels; S3: calculating correlation measurement values of any two channels, and constructing a channel correlation graph; S4: determining a grouping threshold value, and forming an anchor point set from selected anchor point channels; S5: constructing a cross-channel correlation prediction model, and constructing an innovation sequence; S6: determining an error sensitivity coefficient, and determining a quantified error bound; S7: obtaining normalized reconstruction values of the anchor point channels and the innovation sequence; and S8: obtaining normalized reconstruction values of non-anchor point channels, and performing inverse normalization on the normalized reconstruction values according to the scale factor to obtain reconstruction values of the channels under the original dimension. The application can limit compress and highly restore original data.

The invention relates to a knowledge graph redundancy elimination method based on heterogeneous entity attribute semantic alignment and electronic equipment, and the method comprises the steps: inputting two entities of a knowledge graph and a corresponding entity attribute set, and carrying out the preprocessing; calculating attribute similarity according to the attribute types, judging whether the attribute similarity is matched or not, and carrying out weighted summation on each matched attribute similarity value to obtain total similarity; calculating entity name similarity, and calculating comprehensive similarity based on the dynamic weight according to the entity name similarity and the overall similarity; and calculating instance correlation measurement according to the deviation degree of each attribute similarity to the comprehensive similarity and the weight of each attribute in the similarity, and judging whether the two entities are heterogeneous entities of the same object or not to realize redundancy elimination of the knowledge graph. Compared with the prior art, the method has the advantages of being high in accuracy of heterogeneous entity attribute semantic alignment, high in adaptability of rapid scene change and the like.

A method includes accessing a structured content item from a first database and event data from a second database, the event data including sets of event attributes in a multi-dimensional namespace and associated with a respective point in time; determining a relevancy profile characterizing a metric of relevancy of the structured content item over a respective time interval, the metric of relevancy including a distance in the multi-dimensional namespace between attributes associated with the structured content and the sets of event attributes; generating, using the relevancy profile, second digital content including a subset of the structured content item; and providing the second digital content for rendering on a device. Related apparatus, systems, techniques and articles are also described.

The present application is applicable to the technical field of big data, and in particular relates to a sample annotation method and apparatus, a rule relevance measurement method and apparatus, a device, and a medium. The sample annotation method comprises: acquiring a rule set and cleaned relational data; extracting a rule pair from the rule set, and using the rule pair to detect the cleaned relational data to obtain a detection result; and determining the relevance of the rule pair on the basis of the detection result, using the relevance as a label, and associating same with the rule pair to form a rule sample. The rule relevance measurement method comprises: acquiring a predicate vector set which is obtained by performing graph vectorization on relational graph data constructed from all predicates in the relational data; obtaining vectorized rule samples on the basis of the predicate vector set; on the basis of the vectorized rule samples, training a preset measurement model to obtain a trained target measurement model; acquiring a target rule; and on the basis of the target measurement model, obtaining a relevance score value of the target rule. The accuracy of rule relevance measurement is improved.

A method includes accessing a structured content item from a first database and event data from a second database, the event data including sets of event attributes in a multi-dimensional namespace and associated with a respective point in time; determining a relevancy profile characterizing a metric of relevancy of the structured content item over a respective time interval, the metric of relevancy including a distance in the multi-dimensional namespace between attributes associated with the structured content and the sets of event attributes; generating, using the relevancy profile, second digital content including a subset of the structured content item; and providing the second digital content for rendering on a device. Related apparatus, systems, techniques and articles are also described.

The invention relates to the technical field of data mining, and discloses a diversified Top-k rule selection and mining method and system, a terminal and a storage medium, and the method comprises the steps: obtaining a rule correlation measurement function and a rule pair distance measurement function; according to the rule correlation measurement function and the rule pair distance measurement function, an objective function of the rule set is constructed, and the objective function is used for comprehensively evaluating correlation and diversity of the rule set; and generating a Top-k rule set from a rule source according to the objective function, the rule source comprising a candidate rule set or an original data set. According to the method, the correlation and diversity of the rule set can be comprehensively optimized, the rule set with high redundancy is prevented from being output, and the overall quality of the Top-k rule set is improved. Moreover, the Top-k rule can be efficiently selected from the preset candidate rule set, and the Top-k rule can be directly mined from the original data set, so that the practicability is enhanced, and the application range is widened.

This disclosure provides systems, methods, and devices for audio signal processing that support improved keyword detection for speech recognition applications. In one aspect, a method is provided that includes determining a plurality of correlation measures for a series of consecutive audio data frames. Each measure is calculated by obtaining a first feature vector for a respective audio frame and a second feature vector from a preceding frame, then computing the correlation between them. The method further includes identifying the presence of a spoken keyword, determining its start time based on the correlation measures, and defining buffer data for the keyword. Additional aspects are also provided, such as leveraging models to confirm keyword presence, handling background noise, and utilizing circular buffers for efficient processing. Other aspects and features are also claimed and described.

The invention discloses a general text intelligent retrieval method and system based on rule analysis, and relates to the technical field of data processing. The method comprises the steps that feature analysis is conducted on query request data input by a user, and data source features and query intention features are determined; performing multi-dimensional strategy fusion on historical user representation information, the data source features and the query intention features, determining an adaptive retrieval rule and an adaptive evaluation system, and analyzing the adaptive retrieval rule into a structured retrieval condition; after a to-be-retrieved text is segmented, preliminary information retrieval is executed according to the structured retrieval conditions, correlation measurement is conducted on candidate texts according to the adaptation evaluation system, and a semantic matching score set is output; and performing preferential screening and sorting on the candidate texts according to the semantic matching score set, and outputting a structured retrieval result list. According to the method, efficient and accurate semantic retrieval of multiple types of texts is realized.

A method can involve receiving an output signal from each of a plurality of microphones in an environment, each of the plurality of microphones being located at a microphone location of the environment, the output signal corresponding to a person's utterance. The method can involve determining, based at least in part on the output signal, a zone within the environment having at least a threshold probability of including a location of the person, and generating a plurality of spatially-varying attention signals within the zone. Each attention signal can be generated by a device located within the zone. Each attention signal can indicate that a corresponding device is in an operating mode in which the corresponding device is awaiting a command, and can indicate a relevance metric for the corresponding device.

The application belongs to the field of three-dimensional digital measurement, and particularly relates to a point cloud registration method with an attention mechanism, which down-samples input point clouds and target point clouds to facilitate improvement of registration accuracy and efficiency. Spatial and channel attention mechanisms are used to focus on the overlapping area of the two point clouds. The method performs dot product operation on the weighted point cloud features to obtain a correlation matrix corresponding to the points, and then obtains a weighting matrix. The rigid transformation matrix is obtained by singular value decomposition algorithm using the weighting matrix and the two point cloud features, so that the fast and accurate point cloud registration task is realized. The application focuses on the similar areas between the point clouds to be registered by introducing the attention mechanism in deep learning, and extracts the feature information to perform correlation measurement, so as to solve the problem of point cloud registration.

A system and method of processing a search request is provided. Identification of a desired content item is based on comparing a topic of the search request to previous user input. The method includes providing access to a set of content items with metadata that describes the corresponding content items and providing information about previous searches. The method further includes receiving a present input from the user and determining a relatedness measure between the information about the previous searches and an element of the present input. If the relatedness measure is high, the method also includes selecting a subset of content items based on comparing the present input and information about the previous searches with the metadata that describes the subset of content items. Otherwise, the method includes selecting a subset of content items based on comparing the present input with the metadata that describes the subset of content items.

The application relates to the technical field of active metadata management, in particular to an intelligent data management method and system based on active metadata. The method comprises the following steps: calculating the structural complexity of two groups of active metadata according to the number of attributes, the correlation measurement value and the combination entropy value of the two groups of active metadata; calculating the stationary measurement value according to the average value and the maximum value point of a group of active metadata, and obtaining a data characteristic measurement value; further obtaining a change characteristic measurement difference value of the two groups of active metadata; performing weighted summation on the structural complexity and the change characteristic measurement difference value to obtain a complexity difference value; setting the calculation mode of a reference distance of a DTW algorithm, and processing the two groups of active metadata by using the DTW algorithm according to the calculation mode of the reference distance selected according to the complexity difference value of the two groups of active metadata. The application can improve the matching and mapping effect of the active metadata, and realizes the consistency management of multi-source active metadata.

It is proposed a computer-implemented method for selecting values of operating parameters of a machine system by a user, the method comprising the steps of defining a relevance metric for each operating parameter, ranking the operating parameters in an ascending or descending order of relevance metrics, creating at least one user-defined list, the method includes receiving a list of operational parameters, the list including a subset of operational parameters having a highest relevance metric, displaying the user-defined list to a user, receiving a value of the subset of operational parameters of the user-defined list, where the value is provided by the user, and sending the value of the subset of operational parameters to a machine system.

A method, device and computer readable storage medium for reach deduplication across contents is provided. In the method, a plurality of non-overlapping object subsets is determined from an object set based on a plurality of object ranges. An object range is associated with at least one object attribute and targeted by a content of a plurality of contents that targets the object set. A reference number of objects targeted by at least one content in each object subset is determined based on a plurality of individual correlation metrics. An individual correlation metric corresponds to a content and indicates a degree of correlation between objects in the object subset and the content. A target number of objects targeted by the plurality of contents in the object set is determined based on the respective reference numbers in the plurality of object subsets.

Embodiments relate to systems and methods for generating content for a draft message based on tracked product data and pre-processed vectorized data. The methods can include extracting a customer domain identifier from a request and using the customer domain identifier to obtain from a product model a proposed product profile and a proposed product action. The methods can include constructing a first query comprising a semantic-based search criteria and a keyword-based search criteria and obtaining a first set of search results from an internal data store using the first query. The methods can include constructing a second query comprising a search vector and obtaining a second set of search results from a second data store using the second query. The methods can include causing generation of a prompt that includes the proposed product action obtained from the product model, and the composite set of search results satisfying a relevance metric.

The application provides a power grid data security protection method and system, relates to the technical field of data security, and comprises the following steps: constructing a data inference relation graph, recording historical access information of an access subject, and when an access request is received, performing two-dimensional risk assessment based on a derived knowledge set inferred by combination of target data and historical data, a correlation measurement value of historical access and an operating state parameter, and generating an access control instruction. The application can effectively prevent inference attacks based on historical data correlation analysis and improve the power grid data security protection capability.

The invention relates to the field of data mining and machine learning, in particular to a variable correlation analysis method based on machine learning prediction precision, which comprises the following steps: acquiring a data set containing a plurality of variables, and performing data preprocessing; variable pairing and target setting; performing association degree quantification based on a machine learning model; constructing a correlation matrix; and outputting the correlation matrix for visualization or downstream analysis tasks. The method has the advantages that a machine learning model is adopted as a core tool for correlation measurement, any complex nonlinear and non-monotonic relation between variables can be effectively captured, and inherent limitations of traditional correlation analysis methods such as a Pearson coefficient, a mutual information method and a maximum mutual information coefficient are broken through; the abstract correlation is defined as specific'prediction precision ', so that the analysis result has clear and intuitive practical significance, and'the prediction precision of predicting the variable B by adopting the variable A is 0.8' is easy to understand by field experts and be applied to actual decision making.

The invention provides a power grid data security protection method and system, and relates to the technical field of data security, and the method comprises the steps: recording the historical access information of an access subject through constructing a data inference relation graph; and performing two-dimensional risk assessment based on a derivative knowledge set inferred by combining the target data and the historical data and the correlation metric value of the historical access and operation state parameters, and generating an access control instruction. According to the invention, inference attacks based on historical data association analysis can be effectively prevented, and the power grid data security protection capability is improved.