102results about How to "Narrow down the selection" patented technology

The invention provides a deeply differential privacy protection method based on a generative adversarial network. The deeply differential privacy protection method is adopted to solve the problem that attackers use own coding and other methods for restoring training set data when a deep learning model is in use, and by adopting the deeply differential privacy protection method, the purpose of protecting users' privacy in a training dataset is achieved. The method includes the steps that according to the potential dataset scale of the input training dataset, a sensitivity degree and the maximum attacker attacking probability are queried to calculate an upper bound of privacy budgets; during deep network parameter optimizing calculation, a differential privacy thought is integrated, noise data is added, on the basis of the characteristic that differential privacy and Gaussian distribution can be combined, the privacy budget of each layer of a deep network is calculated, and during stochastic gradient descent calculation, Gaussian noise is added to make the overall privacy budget minimum; the generative adversarial network is adopted to generate optimal results which the attackers can get, by comparing attack results with initial data, feedback regulation is conducted on parameters of a deeply differential privacy model, and the balance between dataset availability and privacy protection degrees is achieved.

The semiconductor device 1 includes interconnect layers 10, 20, an IC chip 30, via plugs 42, 44, a seal resin 50, and solder balls 60. The interconnect layer 10 includes a via plug 42. An end face of the via plug 42 on the side of the interconnect layer 20 is smaller in area than the opposite end face, i.e. the end face on the side of the IC chip 30. An end face of the via plug 44 on the side of the interconnect layer 10 is smaller in area than the opposite end face, i.e. the end face on the side of the solder balls 60. The thermal decomposition temperature of the insulating resin 14 constituting the interconnect layer 10 is higher than that of the insulating resin 24 constituting the interconnect layer 20.

The embodiments of the invention provide an education video recommendation method and device, which aims to solve the problem in the prior art that due to the fact that video tags are mixed, disorderly and equally treated, the recommendation accuracy of education videos to users is not high. The method comprises the steps of according to the video attribute information, obtaining a related video set of a target video, wherein the video attribute information comprises a video tag viewing time, a video tag and a video tag category; according to the video tag viewing time, generating a video tag weight, and calculating the similarity between the target video and other targets in the related video set; generating a recommendation list for recommendation according to the similarity between the target video and other targets in the related video set.

Training sample generation is disclosed, Text data, A method for classify public opinion events and relate equipment, in a training sample generation method provided by an embodiment of the present application, At first, that text data is cluster, Because text data is clustered, When a clustering result corresponding to a target category is found, The training samples of the target category can beobtained only by selecting the text data that meet the target category conditions in the corresponding clustering results and then labeling the target category, without analyzing whether the text data in other clusters meet the target category conditions or not. Therefore, the selection range of text data is greatly narrowed, the efficiency of annotation and the accuracy of samples are improved,and the time of annotation text data is shortened. At the same time, it improves the efficiency and accuracy of text data classification and public opinion event classification process.

The invention discloses a method for determining a target communication channel and a server in an embodiment. The server obtains first-type communication channels matched with a task to be executed from N communication channels; under the circumstance that at least one first-type communication channel is an available communication channel, the server determines the target communication channel according to the target score of each first-type communication channel, wherein the target score of each communication channel is obtained by the server at least according to the historical task execution result of the communication channel. In the method for determining the target communication channel, as the target score of each communication channel is obtained according to the historical task execution result of the channel, feedback information is utilized effectively, and therefore, the calculation basis of the target score of each channel is more scientific and comprehensive; in addition, in the process of using the communication channels, the server selects the appropriate communication channel as the target communication channel according to the target score, and therefore, the communication channel can be utilized more reasonably and effectively.

The invention discloses an auxiliary positioning technology for the recognized character used in mobile end yielding OCR function with camera, which comprises the following steps: 1) the operator uses mobile end to shoot the object with character picture information, then it opens the auxiliary positioning function so that the screen displays the tracking cross. The operator moves the cross to let the zero point position inside or on the left angle of the recognized character; at last, it presses the ascertaining key and auto recognizing. 2) The operator opens the mobile end OCR shoot function; the mobile end screen displays a tracking cross. Then the operator adjusts the opposite position to let the zero point position inside or on the left angle of the recognized character; at last, it shoots the picture and auto recognizes it.

The present invention discloses a rapid stock layout method of rectangular workpieces for a single specification board. The rapid stock layout method comprises: (1) a workpiece traversal step of inputting initial data of a plurality of to-be-arranged rectangular workpieces, and traversing the initial data of all the to-be-arranged rectangular workpieces; and (2) a first rectangular workpiece stock layout step of rapidly recommending and arranging a rectangular workpiece as the first rectangular workpiece, arranging the first rectangular workpiece with priority arrangement sequences of left lower, right lower, left upper and right upper, and putting the first rectangular workpiece at a left lower corner, a right lower corner, a left upper corner or a right upper corner of a rectangular single board. According to the rapid stock layout method of rectangular workpieces for the single specification board, a 'clean cut' process requirement is satisfied by a blanking stock layout process, stock layout speed is fast, the total utilization rate of the board is high, manufacturing cost of enterprises is largely reduced, requirements of the enterprises on multispecies concentrated blanking are satisfied, and great benefits are brought for blanking production and selection of the manufacturing enterprises.

The invention relates to a method utilizing lithological combination identification sedimentary facies. The method comprises the following steps that S1, the lithological combination index of a study layer section is collected and processed; S2, the depositional environment is judged utilizing the mudstone color index; S3, sedimentary facies belt is judged utilizing the mudstone thickness index; S4, the material source stability is judged utilizing the sandstone thickness disparity; S5, the study layer section lithology is judged utilizing the lithology index; S6, the hydrodynamic strength is judged utilizing the sandstone grain size index; S7, the grain size variation is judged utilizing the sandstone grain size variation index; S8, the material source stability is judged utilizing the sandstone thickness index; S9, the ranges of the sedimentary facies form step S2 to step S8 is reduced step by step according to the study layer lithological combination index, so that the sedimentary facies is judged; and the step S2 to step S8 are not sequential.

The invention discloses an estimation method of the radius of the buried depth of an underground metal circular pipe based on the BP neural network. The method comprises steps of by use of ground penetrating radar to carry out direct simulation so as to obtain radar echo data of underground metal circular pipes with different radiuses and buried depths in different media; extracting the A-Scan echo peak value right above the metal circular pipes, peak value arrival time and an echo energy value in a preset time period, and based on the three parameters, constructing characteristic vectors so as to a form characteristic vector matrix; using the characteristic vector matrix as a data set of a training sample input part, wherein the data set of the training sample output part consists of the radiuses and the buried depths of the existing metal circular pipes and relative dielectric constants of background media; designing a structure of the BP neural network, and using training sample data to train the BP neural network; after the training is finished, inputting characteristic parameters of to-be-tested ground penetrating radar echo data into the BP neural network; and estimating the radiuses of the buried depths of the underground metal circular pipes. According to the invention, the radiuses of the buried depths of the underground metal circular pipes can be quickly and precisely estimated.

The embodiment of the invention discloses a file management method. The file management method comprises the following steps of judging whether the conditions of triggering file management are currently met or not; if the conditions of triggering the file management are currently met, acquiring file operation information in a current storage space; acquiring target management files according to the file operation information. Correspondingly, the embodiment of the invention also discloses a file management device and a terminal. By the adoption of the file management method, the file management device and the terminal, the target management files are acquired according to the file operation information and can be automatically provided for a user, so that the user can acquire a file needing to be operated from the target management files, a selection range is reduced, the file acquisition efficiency is improved, and the experience of the user is improved.

The invention discloses a book recommendation method based on a sequence pattern. The method comprises the following steps of: sequencing various records in a library borrowing database by taking a library card number as a primary key and borrowing time as a secondary key and transforming the records into a sequence database consisting of borrowing sequence data; solving a large book set of a frequent sequence pattern comprising only one book set in the sequence database, and mapping the large book set to be a group of adjacent integers; replacing various borrowing sequence data in the sequence database by the large book set contained in the sequence data; excavating a frequent sequence pattern set by using the large book set; finally finding out the maximum frequent sequence pattern set in the frequent sequence pattern set to construct a reader borrowing behavior library. Therefore, the reader borrowing behaviors in the library borrowing database can be discovered effectively, and the relevant books can be recommended to readers.

The invention relates to the field of intelligent terminals, in particular to a method and a device for inserting an emoticon in a dialogue interface. The method comprises the steps of starting a camera to capture a current user emoticon when an emoticon input request is obtained; identifying a current emoticon type which the current user emoticon is belonged to; determining and preferentially displaying a preset emoticon corresponding to the current emoticon type. According to the method and the device disclosed by the invention, automatic identification on the current user emoticon and preferential displaying the preset emoticon corresponding to the current emoticon can be realized, and the selection range of emoticons can be lessened, so that convenience is provided for a user to select the emoticon needing to be input; a selection procedure of inserting the emoticon during an information dialogue process can be simplified, better application experience is brought about for the user, and an intelligent terminal can be more fashionable and individual.

The invention belongs to the technical field of wireless communication, and discloses a high-performance indoor positioning method based on a random forest. According to the indoor positioning method,in the process of establishing a position fingerprint database, firstly, access points with poor signal strength and unstable signals are deleted according to the signal strength of the access points; and an access point with a better positioning effect is selected from the remaining access points by using an information gain method to form an access point set. On this basis, a location fingerprint database representing location characteristics is established; and then, position clustering is carried out on all positions in the positioning scene through a clustering algorithm, and then a high-precision and high-stability random forest model is constructed for each position cluster. The position of the user is determined by using a random forest model in the positioning process. By constructing the high-precision and high-stability random forest model, the problems that a single decision tree model is limited in positioning precision, unstable in positioning effect and prone to fallinginto overfitting can be effectively solved, and the positioning stability and positioning precision are improved.

The invention discloses an ultrasonic video image processing method, system and device and a storage medium. The ultrasonic video image processing method comprises the steps of: recognizing standard image frames in an ultrasonic video to be processed; determining all standard video segments in the ultrasonic video to be processed according to the standard image frames; marking a video segment marking frame from the at least two standard image frames contained in each standard video segment; marking a single marked frame; and after associating the video segment marking frame and / or the single-frame marking frame with a progress bar of the ultrasonic video to be processed according to the frame number, displaying the video segment marking frame and / or the single-frame marking frame and the corresponding ultrasonic standard surface type on a preset display interface. According to the method and the device, the video content can be prompted in the display content associated with the progress bar in the preset display interface, so that a user can position the required standard video segment or single-frame marked frame through few operations, the operation difficulty is greatly reduced, the video positioning time is saved, and the positioning efficiency is improved.

The invention relates to a Method and device for creating three-dimensional sketch and computer equipment, and the method comprises the steps: carrying out the feature extraction of a to-be-modified three-dimensional model through the computer equipment, obtaining linear features, carrying out the line segment processing of a two-dimensional stroke input by a user, and determining each straight line segment of the two-dimensional stroke; Then, obtaining linear features of the three-dimensional model and position relations between all straight line segments of the two-dimensional strokes, and generating candidate canvases according to all the position relations and preset canvas forming conditions; And determining a target canvas according to the capability coefficients of the candidate canvases, and performing back projection on the two-dimensional strokes on the target canvas to obtain a three-dimensional sketch. According to the method, the difficulty of creating the three-dimensional sketch can be reduced, so that the creation of the three-dimensional sketch is simpler; Furthermore, the creating efficiency of the three-dimensional sketch can be greatly improved.

The present application discloses a data searching method and system. The method comprises: receiving a first search instruction, wherein a search keyword is carried in the first search instruction; searching according to the search keyword to obtain a first search result and display the first search result; and receiving a second search instruction in a display page of the first search result, and searching according to the search keyword and pre-acquired user behavior data, to obtain a second search result and display the second search result. By means of the method and system disclosed by the present application, a selection range of a user is narrowed down, and search efficiency is increased.

The invention relates to the technical field of engines, and provides an engine thermodynamic cycle parameter analysis method and an engine thermodynamic cycle parameter analysis device. The engine thermodynamic cycle parameter analysis method comprises the following steps: establishing a parameter coupling relationship among all parts of an engine; calculating to obtain an initial thermodynamic cycle parameter selection range according to the parameter coupling relation; determining a strength limiting parameter of each part; and obtaining a target thermodynamic cycle parameter selection range according to the intensity limiting parameter and the initial thermodynamic cycle parameter selection range. According to the engine thermodynamic cycle parameter analysis method, parameters of allparts are mutually coupled, and the thermodynamic cycle parameter selection range obtained through drawing is more accurate; the structural strength limitation of components such as a compressor and aturbine is considered, the thermodynamic cycle parameter selection range is further narrowed, and the performance design reasonability of the turboshaft engine is improved; The number of iterations of the turboshaft engine performance design is effectively reduced, and the project development period is shortened.

A method for selecting transmission format combination of third generation mobile communication system, which contains generating two dimension table of transmission format combination according to configured transmission format combination, selecting out transmission format combination in inactive transmission channel and obtaining a sub table of transmission format combination, inquiring transmission block size, number or buffer area occupying percentage in logic channel mapped by each active transmission channel, determining schedule order, combining sub table using transmission format, selecting transmission format for each active transmission channel in said wireless frame. Said invention priority transmits the data in higher priority logic channel or large buffer area occupied logic channel, reduces buffer area overflowing possibility and greatly raises selection speed.

The invention discloses an intelligent visualization system and method for a constructional engineering construction standard atlas based on Unity 3D. The invention belongs to the field of intelligentbuildings, the technical problem to be solved by the invention is how to convert a traditional 2D paper standard specification atlas into a building engineering construction standard atlas which is three-dimensional, visual, intelligent, interactive and easy to learn. According to the technical scheme, the system comprises a three-dimensional visualization unit and an intelligent interaction unit. According to the method, equal-proportion reduction modeling is carried out on the content described by a planar construction drawing set by utilizing a BIM technology, and colors of different 3D components forming the model are designed and distinguished by referring to a real object; importing the model into a Unity 3D engine, packaging the model into an APK, and exporting the APK to a mobileterminal screen for display, thereby realizing visualization of a construction standard atlas; a UI interface is designed on the basis of a visual interface, free control over the model based on a touch control mobile terminal screen is achieved, and meanwhile ternary information linkage intelligent teaching is achieved based on a touch control screen.

The invention discloses a method and a device for realizing intelligent beam registration by using network environment context information. The method comprises the following steps: calculating transceiving beam pair information corresponding to a communication link for data transmission according to positions of a base station and a mobile terminal, network environment context information and influence information of a surrounding transmission medium on electromagnetic transmission; judging whether effective communication link information corresponding to the corresponding information is stored in a knowledge graph library or not according to the base station position, the mobile terminal position and the weather information; if so, directly selecting transceiving beam pair information corresponding to a communication link for data transmission, otherwise, obtaining real-time transceiving beam pair information through traditional beam training or reconstructing a communication environment stereotactic diagram to calculate switching beam pair information. According to the method, the beam switching position of the mobile terminal is predicted by using an artificial intelligence theory, so that the switching beam pair is pre-estimated. According to the invention, low-overhead training and efficient beam registration can be realized and a communication link can be established ina mobile communication scene.

The invention provides a speech enhancement method / system, a computer readable storage medium and electronic equipment. The speech enhancement method comprises the following steps: performing short-time Fourier transform on the input signal of a microphone array so as to form a frequency domain input signal; generating an optimized super-gain beam former to process the frequency domain input signal through the super-gain beam former so as to form a super-gain beam processing output signal; performing low-frequency white noise gain correction on the super-gain beam processing output signal so as to determine whether the low-frequency white noise gain correction result is the super-gain beam processing output signal or the delayed addition beam processing output signal; and performing low-frequency noise reduction compensation on the low-frequency white noise gain correction result so as to output the frequency domain signal after low-frequency noise reduction compensation. The problem of white noise amplification can be controlled and the low-frequency interference dropping ability can be reduced to a certain extent. Moreover, after application of the speech enhancement method, theselection range of the diagonal loading factor is reduced and the robustness is enhanced so as to be favorable for practical application.

The invention discloses a method and a device for transmitting the information of to-be-executed tasks, which are applied to a server. Multiple communication channels are preset in the server, and the communication channels correspond to the types of tasks. The method is characterized by including the following steps: acquiring to-be-executed task information sent by a client, wherein the to-be-executed task information includes the types of the to-be-executed tasks; determining target communication channels matching the to-be-executed tasks according to the types of the to-be-executed tasks, wherein there is at least one target communication channel; and transmitting the information corresponding to the to-be-executed tasks through the target communication channels. By implementing the embodiment provided by the invention, the processing efficiency of to-be-processed tasks is improved.

The invention discloses a multi-publish-source external routing slave next-hop selection method and equipment. The method comprises the following steps that: after receiving the same external route published by different autonomous system boundary routers (ASBRs), a router calculates master next-hop from the router to the external router and determines a candidate slave next-hop set from the router to the external router, wherein the candidate slave next-hop set consists of the master next-hop and the slave next-hop from the router to the ASBRs which publish the external router or / and the master next-hop and the slave next-hop from the router to a forwarding address (FA) carried in the external route; and the router selects the slave next-hop for the master next-hop from the router to the external router from the candidate slave next-hop set. By adopting the method, aiming at the external route published by multiple sources, the resource expense for calculating the slave next-hop is reduced, and the route constriction speed is increased.

An object of the present invention is to provide a shopping system which allows a user to designate a retrieval priority item and allows a retrieval information service terminal to narrow down information, thereby easily retrieving a target commodity at the time of providing the user a retrieval information list. The present invention is characterized in that latest commodity information is obtained through an information communication network, and a retrieval information list is constituted on a retrieval service terminal so that a user side terminal can acquire the retrieval information list; and based on an information retrieval request relating to a commodity selected from the retrieval information list by the user, the retrieval service terminal collates the commodity information provided by the dealer with the information retrieval request, and refers shopping information narrowed down and thereby specified to the user side terminal.

The invention provides a deeply differential privacy protection method based on a generative adversarial network. The deeply differential privacy protection method is adopted to solve the problem that attackers use own coding and other methods for restoring training set data when a deep learning model is in use, and by adopting the deeply differential privacy protection method, the purpose of protecting users' privacy in a training dataset is achieved. The method includes the steps that according to the potential dataset scale of the input training dataset, a sensitivity degree and the maximum attacker attacking probability are queried to calculate an upper bound of privacy budgets; during deep network parameter optimizing calculation, a differential privacy thought is integrated, noise data is added, on the basis of the characteristic that differential privacy and Gaussian distribution can be combined, the privacy budget of each layer of a deep network is calculated, and during stochastic gradient descent calculation, Gaussian noise is added to make the overall privacy budget minimum; the generative adversarial network is adopted to generate optimal results which the attackers can get, by comparing attack results with initial data, feedback regulation is conducted on parameters of a deeply differential privacy model, and the balance between dataset availability and privacy protection degrees is achieved.

The invention discloses a method for improving call speed and call completion ratio of PHS mobile phone. The PHS mobile phone can first identify a special network condition in a network environment of only one or a few mezzo forte signal base station, and a plurality of weak signal base stations, then use a strategy of incomplete traversing for ignoring multitudinous weak signal base stations, namely, use collecting attempt selected a few base stations in a preset call timer duration, reduce selection range, shorten candidate base station list, avoid low successful probability for establishing link on the weak signal base station and long call time, and avoid conditions of failure for attempting establishing link caused by temporality over weight network load or short time trouble, accordingly, the PHS mobile phone can adapt the present network condition which can solve problem of low call completion ratio and long call completion time caused by the complex weak signal environment.

The invention discloses a power supply method. The power supply method comprises the following steps: receiving application information of a user, wherein the application information comprises an installation address and an installation load requirement; obtaining all power supply point names corresponding to preset limitation conditions according to the installation address; obtaining information of each power supply point from a data acquisition and monitoring control system according to the all power supply point names, wherein the information of the power supply point comprises a rated load and an actual load of the power supply point; and obtaining the installation address and a distance of the power supply point when the installation load requirement is less than a difference value of the rated load and the actual load of the power supply point. The invention further provides a power supply device; and by utilizing the power supply device, the whole handling time limitation of expanding installation business is shortened, and the work efficiency is improved.

In order to solve the technical problems in the prior art, the invention provides an engineering quality acceptance management system and a method based on field positioning. According to the method,an engineering project inspection field is positioned through a positioning base station, and a quality acceptance standard is solidified in a software system, so that a user can check the acceptancestandard conveniently during acceptance, when a quality problem is found, on-site photographing is carried out to propose rectification, and during specific operation, a user only needs to click and select, characters do not need to be input in a time-consuming mode, meanwhile, result input is standardized, numbers are mainly selected and input, the characters are forbidden to be input, and greatconvenience is brought to use of the user; through the engineering project information recommendation selection process, the selection range is reduced, the system operation time of the user is further reduced, the operability of the system is improved, and the existing technical problems are effectively solved.