43results about How to "Shorten the iteration cycle" patented technology

The invention relates to the field of machine learning, and discloses a credit risk assessment model generation method based on machine learning and the related equipment. The method comprises the steps of obtaining loan sample data of a user; counting preset sample data stored in the loan sample data; when the number of preset sample data stored in the loan sample data reaches a preset thresholdvalue, taking the loan sample data as target training data; and converting the target training data into a loss function, and adding the loss function into a preset credit risk assessment model for training, thereby generating a target credit risk assessment model. According to the invention, by pre-judging the loan sample data through the preset sample data; and training the pre-established credit risk assessment model by using the judged effective loan sample data so as to generate a target credit risk assessment model, and training the machine model by using the effective sample data, the iteration period of the machine model is shortened.

The invention provides a retention time violation repairing method, device and equipment. The method includes: before layout time sequence optimization, calculating a retention time violation value ofa scanning time sequence path in an integrated circuit in advance according to a netlist of the integrated circuit; determining a buffer combination required for repairing the retention time violation of the scanning time sequence path according to the retention time violation value of the scanning time sequence path; generating a layout constraint according to the layout space corresponding to the buffer combination, and performing layout time sequence optimization according to the layout constraint to reserve the layout space corresponding to the buffer combination for the scanning time sequence path; during a hold time repair phase, inserting the buffer corresponding to the buffer combination into the layout space corresponding to the reserved buffer combination, and the newly insertedbuffer is located in the reserved layout space, so that the placement of the original unit is not influenced, too many wiring resources are not increased, the influence on the design is minimum, andthe iterative period of the design can be remarkably shortened.

The invention provides an unsupervised pedestrian re-identification method based on category adaptive clustering. The method comprises the following steps: firstly, pre-distributing an initializationlabel for a label-free data set, and extracting high-dimensional features of a pedestrian image by utilizing a CNN model; then, carrying out staged category adaptive clustering on the feature set, updating an image label, calculating an average contour coefficient of a clustering result of the stage, and alternately updating a feature extraction CNN model and a category adaptive clustering model until an inflection point appears in the average contour coefficient; and finally, extracting test pedestrian image features by using the CNN model at the inflection point obtained by training, and calculating a feature distance to obtain an unsupervised pedestrian re-identification result. The method can solve the problem that the number of individuals in an unmarked data set pedestrian re-identification task cannot be determined, can dynamically adapt to data sets of different scales, and has high algorithm efficiency and precision.

The invention provides an obstacle recognition model training method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence. The obstacle recognition model training method comprises the steps of obtaining a newly added sample road image and a pre-trained initial obstacle recognition model, wherein the initial obstacle recognition model comprises a historical detection branch network for detecting a first type of obstacle; determining a second type of obstacle corresponding to the newly added sample road image, and constructing a corresponding target detection branch network for detecting and identifying the second type of obstacle in the initial obstacle identification model; and fixing network weight information corresponding to the historical detection branch network, and training the initial obstacle recognition model according to the newly added sample road image to obtain a trained incremental obstacle recognition model. According to the technical scheme of the embodiment of the invention, in the training process of the incremental obstacle recognition model, the reconstruction of the whole obstacle recognition model and the reduction of the recognition efficiency of old obstacles can be avoided.

The embodiment of the invention relates to the technical field of risk management and control, and discloses an address fuzzy matching method and system fusing multi-dimensional similarity and a ruleset. The method comprises the steps: preprocessing a standard address based on a standard address library to obtain to-be-matched information, extracting the address features of the to-be-matched information, and carrying out the fuzzy matching based on the to-be-matched information and the address features. calculating a first similarity by adopting the similarity calculation model, calculating asecond similarity by adopting the comparison rule set, and adjusting the first similarity by adopting the second similarity to obtain an overall similarity of the standard address relative to the to-be-matched address. According to the embodiment of the invention, a comparison rule set is matched with a similarity calculation model to carry out address fuzzy matching on a standard address, and the fuzzy matching problem of a remote address and an unrecorded address is solved through self-updating of a standard address library and incremental training of the similarity calculation model; and the comparison rule set is synchronously updated according to the incremental training result, manual iteration is not needed, and the iteration efficiency is improved, so that the system maintenance cost is reduced.

The embodiment of the invention provides a mobile software development system and method based on the Web technology, and relates to the technical field of computers. The system comprises a JS interface unit for allowing a developer to call basic functions provided by an iOS SDK and an Android SDK, a bridging library unit for connecting the JS interface unit with a platform frame unit and providing an expansion interface of an assembly, and the platform frame unit for integrating the SDK at the iOS end and the SDK at the Android end and allowing the developer of the corresponding platform to call create project engineering. By means of the unity and operability of the Web technology, the development cycle and project iteration cycle of a mobile APP are shortened.

The invention discloses a special business support system and method based on an enterprise tax system. The method comprises the steps that an information acquisition module receives login informationinput by an access user at a web front end; an authentication module matches the login information with stored account information to obtain a session Id corresponding to the access user and providedby the sso (single sign-on) service, and returns a home page of each subsystem to the web front end through url carrying the session Id, and performs login twice by utilizing the session Id in the url intercepted in the home page when logging in the subsystem so as to realize login of the subsystem; and a data storage module stores the service information configured in the basic platform. According to the invention, Cookie homologous sharing is adopted; single sign-on is realized through a session Id double-layer strategy, the stability of single sign-on is enhanced, high availability is realized by storing user information through a redis cluster, and a service system and a basic data module are decoupled by dividing a workbench module, so that the development workload is reduced, and the development efficiency is improved.

A display and control layout adjustment system comprises a movable bracket mechanism (1) comprising a driving member (11) and a plurality of end effectors (12) connected thereto; A display device (2)includes a plurality of panels, each panel being connected to at least one end effector (12); A manipulating device (3), a driver seat (31) and a pedal (32) of the manipulating device (3) are respectively connected with an end effector (12); A control device (4) for controlling the movement of the movable bracket mechanism (1) according to a layout adjustment command; The movable bracket mechanism(1) drives the display device (2) and the manipulation device (3) to move under the control of the control device (4) for layout adjustment. The layout adjustment system of the present application realizes the quick adjustment of the display and control layout, and at the same time can greatly shorten the iterative period of the cockpit design and reduce the iterative cost.

The invention discloses a video library index table establishing method and device, a server and a storage medium. The method comprises the steps of performing feature extraction on a target video frame of each video in the video library to obtain a feature vector of the target video frame; dividing the feature vector into a plurality of feature sub-vectors corresponding to the plurality of firstsegments according to a preset first dimension division strategy; clustering the feature sub-vectors corresponding to each first segment, and determining a segment center vector of each first segment;determining a Cartesian product between the segmented center vectors to obtain a library center vector set; for any target video frame, determining a target library center vector according to the distance between the target video frame and the library center vector; determining a quantization vector according to the target library center vector; and determining a first index field and a second index field corresponding to the target video frame in the index table according to the target library center vector and the quantization vector. According to the method, the training time consumption and the iteration period of the codebook vector are shortened, and the construction efficiency of the index table is improved.

The invention provides an antenna housing force thermoelectric integrated analysis method, and the method mainly comprises the following steps: S1, establishing a radome mechanical calculation finiteelement model and an electromagnetic calculation undeformed finite element model; S2, carrying out aerodynamic thermal load solving on the mechanical finite element model established in the step S1; S3, carrying out the aerodynamic load solving on the mechanical finite element model established in the step S1; S4, performing post-processing on calculation results of the S2 and the S3, and extracting node displacement; S5, reconstructing a finite element model in S4 based on Python language; S6, reconstructing the finite element model deformed in S5 based on HyperMesh software to obtain an electromagnetic calculation finite element model, and completing the analysis of the wave transmission performance of the radome under the aerodynamic load; S7, completing electromagnetic medium parametersetting and antenna modeling of the electromagnetic calculation finite element model established in the step S1, completing original-state radome wave-transparent performance analysis, and comparingthe original-state radome wave-transparent performance analysis with the calculation result in the step S6.

The invention belongs to the field of casting of high-temperature alloy hollow turbine blades, and relates to a method for self-adaptive positioning of a ceramic mold core in a wax mold pressing mold.By optimizing the structural design scheme of a mold core mold and the wax mold pressing mold, a ceramic mold core formed through sintering is placed in the wax mold pressing mold, and the pose of the ceramic mold core in an appearance mold is controlled through six height-adjustable positioning bosses embedded in a loose piece of the wax mold pressing mold; and it is guaranteed that the mold core and the positioning bosses are in an attached state through 1-3 elastic press-fit devices with a limiting function which are embedded in the loose piece of the wax mold pressing mold. The method issimple and easy to implement, effectively reduces wall thickness out-of-tolerance scrapping caused by core shift of a blade, improves the wall thickness size precision and consistency of a cast part,greatly improves the product percent of pass, reduces economic losses, shortens the product development period and reduces the mold iteration cost.

The invention provides a radome electromagnetic performance analysis method under a force load. The method mainly comprises the following steps: S1, establishing a radome mechanical calculation finiteelement model, and electromagnetically calculating an undeformed finite element model; s2, adding material attributes to the mechanical finite element model established in the step S1, applying aerodynamic load and boundary conditions, and completing solving of a static problem; and S3, carrying out post-processing on the calculation result of S2, and extracting node displacement. S4, reconstructing the finite element model in S3 based on Python language to obtain a deformed finite element model under aerodynamic load; s5, reconstructing the finite element model deformed in the step S4 basedon HyperMesh software to obtain an electromagnetic calculation finite element model, and completing the analysis of the wave transmission performance of the radome under the aerodynamic load; s6, electromagnetic medium parameter setting and antenna modeling of the electromagnetic calculation finite element model established in the step S1.2 are completed, original-state radome wave-transparent performance analysis is completed, and the original-state radome wave-transparent performance analysis is compared with the calculation result in the step S5.

The invention provides a method and a system for anisotropic velocity modeling. The method for the anisotropic velocity modeling comprises performing a first iteration process to obtain an initial inclined lateral isotropic velocity, an initial formation dip angle, and an initial azimuth angle corresponding to the time that a near-offset of a first pre-stack depth migration gather has been leveled; performing a second iteration process to obtain a second inclined lateral isotropic velocity, a first formation dip angle, a first azimuth angle, a first anisotropic parameter, and a second anisotropic parameter corresponding to the time that the near-offset of a third pre-stack depth migration gather has been leveled; performing a third iteration process to obtain a fourth anisotropic parametercorresponding to the time that the near-offset of a fourth pre-stack depth migration gather has been leveled; and obtaining an anisotropic velocity model based on the fourth anisotropic parameter, athird inclined lateral isotropic velocity, a second formation dip angle, a second azimuth angle, and a third anisotropic parameter. According to the method and the system for the anisotropic velocitymodeling, the iteration cycle can be shortened; the accuracy of the offset imaging can be improved; and the error between the detected data and the real data can be reduced.

The invention discloses a UGC mode-driven UI visualization system, which comprises a user-oriented visualization platform constructed at the front end of a network, and a material library and a creative template library constructed at the background of the network, and is characterized in that the material library stores a plurality of bottom layer elements capable of forming a UI interface in a data manner; wherein various types of UI visual interface creative templates are stored in the creative template library; the visualization platform comprises an example control area, a creativity display area and a target layout area. Example controls in the example control area map bottom-layer elements in the material library in a one-to-one correspondence mode, the target layout area is in datatransmission association with the creative display area, the example control area, the material library and the creative template library of the network front end at the same time, and the creative display area is in data transmission association with the creative template library of the network background. According to the method, the UGC mode is applied to the UI visual interface platform, andpopularization and efficiency improvement of enterprise-level user experience design visual application are driven.

The invention provides a unified workflow implementation method, which comprises the following steps: S1, a unified workflow platform is built based on an extensible cloud architecture and a businessprocess management engine, and configuration management is performed on the unified workflow platform; s2, the unified workflow platform adopts a process virtual machine, integrates external servicesthrough a service interface, and provides a graphical interface for an administrator user to realize docking between the administrator user and the services; s3, the administrator user acquires the service according to the graphical interface, creates a flow model according to the service, simulates the flow model, publishes the flow model passing the simulation, and provides the flow model to anexternal system user for use to generate a flow instance; and S4, the unified workflow platform acquires a flow model created by the administrator user according to the service and a flow instance generated by the system user using the flow model, and manages the flow model and the flow instance.

The invention relates to an unsupervised pedestrian re-identification method based on hierarchical clustering and a difficult sample triple. The method comprises the following steps: 1) dividing a pedestrian image set into a training set and a test set; 2) initializing a label; 3) inputting the training set into a CNN model, extracting pedestrian features, and performing model training; 4) carrying out hierarchical clustering on the pedestrian features, and distributing the pedestrian features into clusters; 5) performing sampling according to the cluster category to obtain a new training set; 6) according to a difficult sample triple loss function, performing fine adjustment on the new training set; 7) according to the average contour coefficient of the training set, performing judging to return to the step 3) or judge the optimal CNN model; and 9) loading the test set into the optimal CNN model to obtain a pedestrian re-identification result. Compared with the prior art, the method not only can solve the problem that the number of individuals cannot be determined in pedestrian re-identification of a no-label data set, but also reduces the probability that different types of highly similar samples are clustered into one type, so that the identification performance of the model is greatly improved.

The invention discloses a service monitoring method and system based on a cloud native data warehouse, and relates to the technical field of the Internet. The method comprises the steps: synchronizing data generated in each service field to the cloud native data warehouse, querying corresponding data from the cloud native data warehouse based on preset index calculation configuration data, and carrying out the index calculation to generate index data; and calculating the index data according to a preset alarm rule to obtain alarm information, and notifying an alarm receiver when the alarm information reaches a preset alarm threshold value. Compared with a traditional method, the service monitoring method and monitoring system based on the cloud native data warehouse solve the problems that the iteration period of incremental indexes is long, index data are dispersed, and the real-time performance of index calculation cannot be guaranteed.

The invention provides an X-ray automatic detection device and method for detecting the quality of a transformer coil. The device comprises a rotating platform, a base, a coil tray, bamboo joint typeelectric hanging brackets, an X-ray generator, an X-ray receiver, and a portal frame.The rotating platform is fixed to the ground, the base is arranged at the top of the rotating platform, the coil tray is arranged at the top of the base, the number of the bamboo joint type electric hanging brackets is two, the two bamboo joint type electric hanging brackets are respectively arranged on a beam ofthe portal frame, and the X-ray generator and the X-ray receiver are fixed at the bottoms of the two bamboo joint type electric hanging brackets respectively. The device is high in automation degree,long in continuous detection time and high in detection efficiency, manpower is greatly saved, the detection result is accurate, and the outgoing quality of the transformer coil is guaranteed.

The invention discloses a multi-target optimization robust collaborative design method for a transmission chain of a low-wind-speed wind turbine generator, and belongs to the field of wind power generation. According to the technical scheme, the method is characterized by comprising the steps: 1, formulating evaluation standards of a low-wind-speed wind turbine generator transmission chain system;(2) formulating selection indexes and evaluation factors related to a transmission chain and components; 3, setting the weight of each design evaluation factor; 4, employing experts and setting influence weights; 5, scoring each scheme by each expert; 6, calculating a weighted average value of each scheme; 7, performing comparing to obtain a transmission chain design scheme with the highest score; and 8, obtaining an optimal low-wind-speed wind turbine generator transmission chain design scheme. According to the method, the problems that a low-wind-speed wind turbine generator has higher requirements on the transmission chain and the optimal design scheme of the transmission chain is not easy to determine are solved.

The invention provides a radome electromagnetic performance analysis method considering a heat effect. The method mainly comprises the following steps: S1, establishing a radome mechanical calculationelectromagnetism calculation finite element model; s2, adding material attributes to the mechanical finite element model established in the step S1, and applying a temperature load and boundary conditions; s3, solving the static problem, and extracting node displacement; s4, reconstructing a finite element model of deformation under a temperature load based on the node displacement extracted in S3; s5, reconstructing the finite element model deformed in the step S4 based on HyperMesh software to obtain an electromagnetic calculation finite element model; s6, electromagnetic medium parameter setting and antenna modeling of the electromagnetic calculation finite element model established in the S5 are completed, and analysis of the wave-transparent performance of the radome under the temperature load is completed; and S7, electromagnetic medium parameter setting and antenna modeling of the electromagnetic calculation finite element model established in the step S1.2 are completed, and original state radome wave transmission performance analysis is completed.

The invention belongs to the technical field of machine vision detection, and in particular relates to a method for generating an irregular hole shape defect image, comprising: selecting a defect-free image as a background image, randomly generating two closed irregular circles on the background image, two non-defective The regular circles are nested, and the two irregular circles are filled with different colors. This method uses an algorithm to randomly and quickly generate a large number of defect samples in the form of holes, which can quickly improve the detection effect of the deep learning model in a short period of time and greatly reduce the iteration cycle of the algorithm.

The invention provides a radome electromagnetic performance analysis method considering a heat effect. The method mainly comprises the following steps: S1, establishing a radome mechanical calculationelectromagnetism calculation finite element model; s2, adding material attributes to the mechanical finite element model established in the step S1, and applying a temperature load and boundary conditions; s3, solving the static problem, and extracting node displacement; s4, reconstructing a finite element model of deformation under a temperature load based on the node displacement extracted in S3; s5, reconstructing the finite element model deformed in the step S4 based on HyperMesh software to obtain an electromagnetic calculation finite element model; s6, electromagnetic medium parameter setting and antenna modeling of the electromagnetic calculation finite element model established in the S5 are completed, and analysis of the wave-transparent performance of the radome under the temperature load is completed; and S7, electromagnetic medium parameter setting and antenna modeling of the electromagnetic calculation finite element model established in the step S1.2 are completed, and original state radome wave transmission performance analysis is completed.

The invention belongs to the technical field of machine vision detection, and particularly relates to an irregular hole morphological defect image generation method, which comprises the following steps: selecting a defect-free picture as a background image, randomly generating two closed irregular circles on the background image, nesting the two irregular circles, and filling the two irregular circles with different colors. According to the method, a large number of hole-shaped defect samples are randomly and rapidly generated by adopting an algorithm, the detection effect of a deep learning model can be rapidly improved in a short time, and the iteration period of the algorithm is greatly shortened.

The invention provides a named entity recognition system based on a deep network ASLSTM. The system comprises a network model BIASLSTMCRF, wherein the network model BIASLSTMCRF comprises a text feature layer, a context feature layer BIASLSTM, and a CRF layer, the context feature layer BIASLSTM comprises two ASLSTM deep networks, and the two ASLSTM deep networks are spliced to form a bidirectionalASLSTM network. According to the method, a novel ASLSTM deep network is designed in the named entity recognition system, the more stable and accurate cell state of the preamble and the postamble of the named entity in the input text can be obtained, and the network depends on self-learning, so context-related semantic representation can be learned, robustness for coping with the unrelated words ofthe precedent and the consequent can be improved, and errors of an identification system are reduced.

The embodiment of the invention provides a method and device for generating a UI interface, and belongs to the field of mobile terminal development. The method comprises the steps that under the condition that an instruction for opening a UI interface is received, service data corresponding to the UI interface and the name of an XML template file are collected; obtaining an XML template file corresponding to the name; replacing a display form of a component in the obtained XML template file with a corresponding preset display form of the component in an application program to which the UI belongs, and obtaining an updated XML template file; and rendering the service data on the updated XML template file to obtain the UI interface. Therefore, real-time updating is achieved, the iteration period is shortened, the verification and trial and error cost is reduced, and the original rendering performance is improved.

The invention discloses a service system management method and device, and relates to the field of computers. A specific embodiment of the method comprises the following steps: receiving a management request, wherein the management request indicates a to-be-managed service and a management type in a service system; determining a first exception code of the to-be-managed service, an upstream service capable of calling the to-be-managed service and a second exception code of the upstream service in response to the fact that the management type is a newly-added type; determining a system code segment corresponding to the to-be-managed service, wherein the number of abnormal codes corresponding to the system code segment is greater than or equal to the number of second abnormal codes; determining system exception codes corresponding to the to-be-managed service from the system code segment, wherein the number of the system exception codes is smaller than or equal to the number of the first exception codes; and according to a corresponding relationship between the system code segment and the second abnormal code, constructing a mapping relationship among the system abnormal code, the first abnormal code and the second abnormal code so as to respond to the management request. According to the embodiment, by establishing the mapping relation, code modification is avoided, and the development efficiency is improved.

The invention discloses a film hole grid modification method, relates to the technical field of computational fluid dynamics application, solves the problems that manpower and time resources are wasted and the speed of optimization iteration is limited in an existing grid modification method, and is realized by a grid analysis module, a parameter modification module, a quantity modification module and a curved surface modification module. The grid analysis module reads the existing grid file, analyzes the position information of the target gas film hole, and obtains the parameters of the target gas film hole and the information of surrounding grid points; the parameter modification module modifies the aperture, the inclination angle and the position so as to modify the parameters of the film hole; a number modification module is adopted to modify the number of the target gas film holes after parameter modification, and deletion and increase of the gas film holes are achieved; and the curved surface correction module corrects the grid points near the target gas film hole according to the coordinates of the points of the wall surface near the gas film hole and the fitting, so that the shape of the wall surface near the modified gas film hole is not changed. The complexity of the method is reduced, and the universality of the method is improved.

The invention discloses a community discovery method based on an iterative greedy thought, and belongs to the field of complex network analysis. The method includes the specific steps that 1, communities are quickly merged based on the greedy thought so as to construct initial community partition; 2, a neighborhood of an initial solution is searched for a better solution through local search to further improve the solution quality; 3, the local optimal solution is disturbed, and nodes with a low contribution to the modularity value are initialized; 4, the distributed solution is reconstructed;5, a neighborhood of the reconstructed solution is searched for a better solution through local search to further improve the solution quality; 6, whether the newly generated solution is used in thenext iteration process or not is judged according to an acceptance criterion; 7, whether a stop condition is met or not is judged. The method is reasonable in design, has a strong global search capability, and can quickly and effectively obtain high-quality community division.

The present invention proposes a radome force thermoelectric integration analysis method, the method mainly includes the following steps: S1: establish the finite element model of the mechanical calculation of the radome, and the undeformed finite element model of the electromagnetic calculation; S2: the mechanical finite element model established for S1 The element model is used to solve the aerodynamic and thermal loads; S3: the mechanical finite element model established by S1 is used to solve the aerodynamic loads; S4: the calculation results of S2 and S3 are post-processed to extract the node displacement. S5: Complete reconstruction of the S4 finite element model based on Python language; S6: Reconstruct the deformed finite element model in S5 based on HyperMesh software to obtain the electromagnetic calculation finite element model, and complete the analysis of the wave transmission performance of the radome under aerodynamic load; S7 : Complete the electromagnetic medium parameter setting and antenna modeling of the electromagnetic calculation finite element model established in S1, complete the wave transmission performance analysis of the original state radome, and compare with the calculation results of S6.

The invention discloses a method, device, server and storage medium for establishing an index table of a video library. The method includes: performing feature extraction on a target video frame of each video in the video library to obtain a feature vector of the target video frame; The first dimension division strategy divides the feature vector into multiple feature subvectors corresponding to multiple first segments; clusters the feature subvectors corresponding to each first segment, and determines the segment center of each first segment Vector; determine the Cartesian product between each segment center vector to obtain the library center vector set; for any target video frame, determine the target library center vector according to the distance between it and the library center vector; determine the quantization according to the target library center vector Vector; according to the target library center vector and the quantization vector, determine the first index field and the second index field corresponding to the target video frame in the index table. The invention shortens the training time and iteration period of the codebook vector, and improves the construction efficiency of the index table.