10 results about "Matched sample" patented technology

The present application relates to an entity matching method, device and medium based on meta-rule induction and retrieval, which comprises: obtaining a label data set containing matching / mismatching label pairs; obtaining a representative sample subset by using a sampling method based on label balance and semantic diversity; for each label pair in the subset, generating matching reasoning basis and checking and correcting by a large language model, and refining into an instance-level natural language matching rule; embedding the instance-level rule into a semantic vector and clustering, each cluster synthesizes a meta-rule to form a rule base; rewriting the meta-rule into a task-adaptive rule according to the target scene; encoding the to-be-matched sample pair into a query vector, calculating the semantic similarity with the rule vector, and retrieving K rules by the maximum edge correlation criterion; inputting the to-be-matched sample pair and the retrieved rule into the large language model, and outputting the matching label. The present application refines the scarce supervision into the retrievable rewritten meta-rule, and realizes the stable generalization of the cross-entity matching task.

This invention provides a method, apparatus, device, and storage medium for inferring a single-sample ncRNA regulatory network, relating to the field of gene identification technology. The specific implementation includes: acquiring ncRNA and target gene transcriptome data of a matched sample; for each target sample in the matched sample, calculating a first regulatory relationship strength matrix and a second regulatory relationship strength matrix of ncRNA and target gene before and after removing the target sample using a preset statistical algorithm; obtaining the fusion regulatory relationship strength matrix of ncRNA and target gene corresponding to the target sample based on the first and second regulatory relationship strength matrices; and inferring the single-sample ncRNA regulatory network of the target sample based on the fusion regulatory relationship strength matrix of ncRNA and target gene corresponding to the target sample. This invention can reflect the regulatory relationship strength between ncRNA and target gene at the single-sample level by constructing statistical correlation values, thereby inferring the single-sample ncRNA regulatory network.

The invention provides a dynamic feature selection and distribution method based on mass score distribution, belongs to the technical field of label distribution, and aims to solve the problem of poor training effect caused by inaccurate sample distribution in the training process of a target detection model. Adopting an intersection-to-union ratio as measurement, constructing a matching matrix of the sample and the label, and calculating a matching score; s2, performing descending order arrangement on each labeled matching sample according to the matching scores, and calculating the matching score difference of adjacent samples in the arrangement; s3, performing positive sample selection on each gt sample; and S4, calculating classification loss and positioning regression loss of the target detection model according to the selected positive sample, and adding the classification loss and the positioning regression loss to obtain final loss.

The application relates to the technical field of information retrieval, and in particular to a retrieval result determination method and device, electronic equipment and a storage medium. The method comprises the following steps: determining first association feature information of a target retrieval word and a candidate object based on semantic information of the target retrieval word and semantic information of the candidate object; inputting the first association feature information into a matching data prediction network to obtain candidate matching data of the target retrieval word and the candidate object; the matching data prediction network is obtained by training based on data matching samples containing sample label data; the sample label data is obtained by performing information matching on sample retrieval words and sample objects; and a plurality of target retrieval objects corresponding to the target retrieval word are determined from the candidate object based on the candidate matching data. The application can improve the accuracy of sample construction and the accuracy of retrieval results.

The invention discloses an address data processing method and a geocoding method, and the address data processing method comprises the steps: processing sample data, and obtaining partial order relation sample data for sample target address data; according to the preset correlation index data and correlation index data between multiple pieces of specified sample candidate address data contained in the partial order relation sample data and the sample target address data, obtaining matched sample candidate address data of which the correlation index data with the sample target address data is greater than the preset correlation index data. The initial model is trained through the partial order relation sample data, so that the association index data between the target recall address data predicted by the trained address association index model in the multiple recall address data of the to-be-queried address data and the to-be-queried address data is greater than the preset association index data; and the accuracy of the target recall address data obtained by the model is improved.

The invention relates to the technical field of fault monitoring, in particular to an oil filter blockage state recognition method based on Internet of Things data, which comprises the following steps: acquiring an operation parameter sample and an environment parameter sample of an oil filter; first clustering is conducted on all the operation parameter samples, a plurality of class clusters are obtained, the class with the most sample labels in the class clusters serves as the matched sample label of each operation parameter sample in the class clusters, and the sample labels indicate the blocking state of the engine oil filter; determining initial label deviation values of the unmatched samples based on the sample labels of the matched samples and other unmatched samples in the class cluster, the matched sample labels and the corresponding environment parameter samples; and taking the initial label deviation value as an initial weight of a to-be-trained blockage state recognition model, training the to-be-trained blockage state recognition model to obtain a trained blockage state recognition model, and recognizing the blockage state of the to-be-detected oil filter by using the trained blockage state recognition model.

The present disclosure provides a data processing model training method and device, and a data processing method and device, relates to the technical field of data processing, and in particular to large model technology and content recommendation technology. The implementation scheme is as follows: determining a training sample set, wherein each training sample in the training sample set comprises spliced sample content and a review text for the sample content; determining at least one matching sample matched with the training sample; encoding the training sample by using a data processing model having current parameters to obtain a representation vector of the training sample; encoding the at least one matching sample by using the data processing model having the current parameters to obtain a representation vector of the matching sample; and adjusting the current parameters of the data processing model to minimize the loss between the representation vector of the training sample and the representation vector of the matching sample.

The invention discloses a face recognition method and device, equipment and a storage medium, and relates to the technical field of machine learning, and the method comprises the steps: collecting historical face data samples to obtain a face information sample set, and carrying out the clustering processing of the face information sample set through a clustering algorithm, so as to obtain a plurality of sample clusters; respectively determining farthest adjacent samples in each sample cluster, and constructing an outlier sample set and each normal sample set; obtaining to-be-recognized face data to obtain a current to-be-matched sample, determining a first distance between the current to-be-matched sample and each second sample in the outlier sample set by using a preset processing unit, and determining an initial nearest neighbor search result of the current to-be-matched sample from each normal sample set by using the preset processing unit; and determining a target nearest neighbor search result corresponding to the current to-be-matched sample based on the first distance and the initial nearest neighbor search result, and determining an identification result of the to-be-identified face data based on the target nearest neighbor search result. In this way, the face recognition efficiency can be improved.

The invention discloses a training method of a semantic analysis model for recalled positions and resumes. The training method of the semantic analysis model used for recalled positions and resumes comprises the steps that delivery behavior data, user resume data and position description data of a user are obtained, and the delivery behavior data comprise positions where the user delivers the resumes; based on the positions of the resumes delivered by the user, constructing matching samples of the positions of the resumes; based on a pre-configured language analysis model, identifying the delivery behavior data, the user resume data and the position description data corresponding to the matching samples, and screening out the matching samples corresponding to the user resume data which does not meet the position constraint requirements to obtain positive matching samples; in the samples which do not meet the position constraint requirements, negative matching samples are extracted; and performing fine tuning training on a pre-constructed semantic analysis model based on the positive matching sample and the negative matching sample to generate a semantic analysis model of the position and the resume.

The application provides a data matching method and a data matching device, which are applied to the field of data processing and are beneficial to improving data processing efficiency. The method comprises the following steps: matching sample data by using a first matching rule to obtain a first matching result; determining a first evaluation index based on the first matching result; adjusting a preset field of the first matching rule to obtain a second matching rule in the case that the first evaluation index does not satisfy a preset condition, and the adjustment comprises at least one of the following: adding the preset field, deleting the preset field, changing a matching range of the preset field or changing matching content of the preset field; matching the sample data by using the second matching rule to obtain a second matching result; determining a second evaluation index based on the second matching result; and matching to-be-matched data by using the second matching rule in the case that the second evaluation index satisfies the preset condition.