Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

9 results about "Global statistics" patented technology

Method for disambiguating knowledge points in text knowledge base

The invention belongs to the technical field of text disambiguation, and particularly relates to a method for disambiguating knowledge points in a text knowledge base, which comprises the following steps of: acquiring text content, extracting knowledge points to be disambiguated from the text content through a knowledge base, and preprocessing the text content; performing global statistics on the knowledge points to be disambiguated, converting the knowledge points to be disambiguated into vectors to be verified, and storing the vectors to be verified into a vector database; generating a plurality of candidate knowledge points according to the knowledge points to be disambiguated; and calculating the similarity between the knowledge points to be disambiguated and the candidate knowledge points through the disambiguation model, and obtaining a similarity comprehensive score of each candidate knowledge point. According to the method, the knowledge points to be disambiguated are obtained, the candidate knowledge point set is obtained through community detection and deep learning, the similarity score is calculated from multiple dimensions such as semantics and context, and the optimal disambiguation result is determined. Compared with a traditional method, the problem of inaccurate disambiguation caused by insufficient semantic understanding is avoided, and accuracy, reliability, integrity and efficiency are improved.
Owner:BEIJING HUAYUN WORLD TECH CO LTD

Page usage behavior data statistics method, device and storage medium based on Flutter without embedding points

The present invention provides a method, device and storage medium for non-buried point statistics of page usage behavior data based on Flutter. By passing the global listener instance object of the page life cycle to the page state abstract class, when the page life cycle changes, the page state abstract class calls back the four interface methods corresponding to the instance object, and obtains the page name indicated by the parameters in the four interface methods to upload page statistics data, so there is no need to write cumbersome buried point codes. The present invention takes over all component pages uniformly at the system level and exposes the global statistics interface to developers, so there is no need to write buried point statistics code for each page component, reducing duplication of work, avoiding omissions and errors, and greatly improving development efficiency.
Owner:GUANGRUI TECH (SHENZHEN) CO LTD

Global heap size coordination of garbage collected workloads

Mechanisms are provided for dynamically modifying maximum heap sizes of local memory heaps. A plurality of applications execute that each have a corresponding local garbage collector in a plurality of local garbage collectors. A global garbage collection manager executes and communicates with the local garbage collectors. The global garbage collection manager maintains global statistics information that includes statistics information communicated to it from each of the local garbage collectors. A first local garbage collector, associated with a first application, communicates first statistics information of the first application, in response to a first local garbage collection operation executed by the first local garbage collector, to the global garbage collection manager. The global garbage collection manager dynamically modifies a heap size associated with the first application based on the first statistics information and the global statistics information.
Owner:ORACLE INT CORP

System and method for determining object cardinality in object store database systems

In a database system, wherein data is stored as objects within an object storage system, a system and method for estimating object cardinality, determining query execution plan costs, and selecting a query plan for execution by the database system. Multiple object cardinality estimation approaches for estimating the number of objects to be accessed for a given query condition on a column of a relation composed of a set of objects, where each object maintains the minimum value and the maximum value of individual columns are presented. A set of global statistics is also maintained, consisting of the total number of objects and the minimum and maximum values of individual columns. The object cardinality estimation is determined based on the global statistics without retrieving individual object-level statistics.
Owner:TERADATA US INC

Method for predicting euploidy based on blastocyst development kinetics and spherical harmonic decomposition fusion

PendingCN122290107AAchieve multi-scale quantificationeliminate distractionsTrophoblastSpherical harmonic analysis
This invention discloses an euploidy prediction method based on the fusion of blastocyst developmental dynamics and spherical harmonic decomposition. The method acquires multifocal plane image sequences of blastocysts using a time-difference imaging system and extracts dynamic parameters. After semantic segmentation and depth estimation, a three-dimensional surface model is obtained through surface reconstruction. The trophoblast cell instances are segmented to obtain three-dimensional centroid coordinates. These centroid coordinates are radially projected onto a unit sphere, and a spherical density function is constructed after excluding the inner cell mass mask. Spherical harmonic decomposition is then normalized using Monte Carlo zero-model normalization to obtain normalized power spectra at each degree. The basic morphological features, dynamic parameters, and spherical harmonic features are fused and filtered before training an ensemble learning model to output prediction results. This invention is the first to introduce spherical harmonic analysis into blastocyst assessment, capturing multi-scale spatial distribution information that global statistics cannot obtain. The power spectrum exhibits rotational invariance, and zero-model normalization eliminates cell number confounding.
Owner:TONGJI HOSPITAL ATTACHED TO TONGJI MEDICAL COLLEGE HUAZHONG SCI TECH

Two-stage speculative decoding optimization method and system for large model inference acceleration

PendingCN122452635AData setAlgorithm
The application discloses a two-stage speculative decoding optimization method and system for large model inference acceleration. First, a general domain dataset and a specific task domain distillation dataset are constructed, and a frozen verification model and a candidate generation model to be trained are established. Based on the general domain dataset, the first-stage preliminary capability training of the candidate generation model is performed, so that the candidate generation model enters the initial interval effectively aligned with the verification model. Based on the task domain distillation data, the second-stage key mark optimization training is performed. The ratio of the verification model probability to the candidate generation model probability is calculated for each to-be-predicted mark to obtain the potential contribution index of the candidate generation model. The dynamic screening threshold and the sparse mask are generated by using global statistics, and the loss optimization is only performed on the key contribution mark. Finally, the trained candidate generation model and the verification model are used to perform speculative decoding inference. The method improves the acceptance length of the candidate mark and the inference acceleration effect, and reduces the training cost.
Owner:SOUTHEAST UNIV

Debugging bus supporting high-frequency signals

The invention discloses a debugging bus supporting high-frequency signals, and belongs to the field of communication. Based on a traditional chip debugging bus, the traditional chip debugging bus comprises a multiplexer, a chip pin and a plurality of modules. The plurality of modules are connected with the multiplexer, and the multiplexer is connected with the chip pins; the debugging bus supporting the high-frequency signals further comprises a global statistical module. The statistical module is positioned between the multiplexer and the chip pin; data in a chip and a clock to which the data belong are input into the statistical module together, and the statistical module detects high-frequency signals and meets statistical requirements of all modules. According to the invention, high-frequency signals can be supported, data statistics can be flexibly carried out on all modules, and meanwhile, fewer hardware resources are consumed.
Owner:上海芯炽科技集团有限公司

System and method for determining object cardinality in object store database systems

In a database system, wherein data is stored as objects within an object storage system, a system and method for estimating object cardinality, determining query execution plan costs, and selecting a query plan for execution by the database system. Multiple object cardinality estimation approaches for estimating the number of objects to be accessed for a given query condition on a column of a relation composed of a set of objects, where each object maintains the minimum value and the maximum value of individual columns are presented. A set of global statistics is also maintained, consisting of the total number of objects and the minimum and maximum values of individual columns. The object cardinality estimation is determined based on the global statistics without retrieving individual object-level statistics.
Owner:TERADATA US INC

Large model generation Chinese text detection method and system based on lexical probability statistics

The invention discloses a large model generation Chinese text detection method and system based on lexical probability statistics. The method comprises the following steps: firstly, calling a proxy large language model to reasone a target text to obtain the logarithmic probability of each token, and then respectively calculating global statistics and local statistics based on a token logarithmic probability sequence of the target text; and finally, dividing the global statistical magnitude by the local statistical magnitude to obtain a detection score of the target text, and comparing the detection score with a given threshold value to make a decision. According to the method, the global and local statistics of the token logarithmic probability sequence of the target text are combined, and low-cost and high-performance large language model generation text detection is realized. Compared with an existing detection method only utilizing global statistics, the method has the advantage that higher detection precision is realized in complex scenes such as cross-model and cross-domain scenes.
Owner:ZHEJIANG UNIV +1