212results about How to "Speed up iteration" patented technology

The invention discloses a page processing method and device of mobile applications, and in particular to a method and a device for updating client-side pages of the mobile applications and a page processing system of the mobile applications. The page processing method of the mobile applications mainly comprises the steps of analyzing pages returned by a server side and transforming the pages into objects which can be recognized by a client side; traversing each page component in the objects which can be recognized by the client side and generating examples of the page components according to types of the page components and a base component library of the client side; configuring the examples of the page components according to attributes of the page components; rendering the examples of the page components; returning all the rendered examples of the page components to the objects which can be recognized by the client side, generating and displaying local objects of the client side of the pages. By adopting the method provided by the invention, the mobile applications can have the advantages of low development cost like web applications and good user experience like native applications.

The invention discloses a parallel deep learning scheduling training method and system based on a container, and belongs to the technical field of cloud computing and deep learning. The technical problem to be solved by the invention is how to avoid that each Task resource of TensorFlow cannot be isolated during training, the mutual influence is caused by resource preemption, the defect schedulingcapability, the upper-layer development amount is large, and the checking of each Task training task and log is inconvenient. The adopted technical scheme is as follows a Kubernetes container is utilized to realize the configuration and scheduling of the computing resources of tasks, a plurality of resource management mechanisms such as ResponceQuota and LimitRanger are provided, and the resourceisolation among the tasks is realized through communication among pod nodes in a container cluster; the same training node starts training pod and life cycle management pod at the same time, the LCMcarries out resource job scheduling in a unified mode, and the micro-service framework serves as POD deployment and depends on the latest version characteristic of Kubernetes to effectively mobilize the use of the GPU. The invention also discloses a parallel deep learning scheduling training system based on the container.

A method to incorporate the steady-state model of the generalized power flow controller into a Newton-Raphson power flow algorithm is disclosed. The disclosed method adopts a flexible steady-state model of the generalized power flow controller, which can be applied to calculate the power flow solution of a power grid embedded with STATCOM, UPFC, GUPFC and the generalized power flow controller in a single framework. The disclosed method only incorporates the control variables of the shunt voltage sourced converter into the state vector of Newton-Raphson power flow algorithm. The increment of state variables due to incorporating the generalized power flow controller is less than the prior art. Further, the method can preserve the quadratic convergence characteristic of the Newton-Raphson power flow algorithm after embedding the generalized power flow controller into a power grid.

A first processor generates a private crypto-key and a public crypto-key. The first processor divides the private crypto-key into two portions, a first private key portion, based upon a user's password, and a second private key portion. The private crypto-key and the first private key portion are then destroyed. The remaining portion, second private key portion, and the public crypto-key are stored in a memory. A second processor generates the first private key portion based upon the user's password and responsive to receiving the user's password. The second processor then destroys the generated first private key portion with out storing the generated first private key portion.

Embodiments described herein may provide devices, systems, methods, and / or computer readable medium for adverse event detection and severity estimation in surgical videos. The system can train multiple models for adverse detection and severity estimation. The system can load selected models for real-time adverse event detection and severity estimation.

The invention discloses an autonomous collision avoidance decision-making method for an unmanned ship based on adaptive navigation situation learning. The autonomous collision avoidance decision-making method comprises the steps that firstly, navigation state information of the unmanned ship is analyzed and described, and a navigation situation estimation body concept model of entity class and seaarea attribute in the navigation environment is established; secondly, a relation between the unmanned ship and an obstacle is determined as a binary relation, the body model is quantitatively divided into various navigation situation sub-scenes by combining with the international maritime collision avoidance rule; and thirdly, the current environmental state information of the unmanned ship in the sub-scenes is obtained, a feedback memory unit of a long-short term memory network is constructed, a ship autonomous collision avoidance decision-making algorithm is utilized to interact with the marine environment, and the optimal strategy of autonomous collision avoidance is calculated through adaptive navigation situation learning. According to the autonomous collision avoidance decision-making method, dimensionality reduction is conducted on the navigation situation of collision avoidance decision-making adaptive learning, thus the feasibility of decision-making is greatly improved, theiterative speed of the algorithm is greatly increased, and real-time autonomous obstacle avoidance and navigation safety of the unmanned cargo ship are ensured.

The invention discloses a method and device for achieving a low delay basic transcendental function based on a mixed model CORDIC algorithm. The method comprises the following steps that the angle and function type in the floating point form in the IEEE-754 standard is input, format conversion and compressed mapping are conducted; rotating direction prediction is conducted on a Z data path according to a mapping angle and the function type; the X data path and the Y data path divide 64 iterations into the front 32 eight-level compressed iterations through a carry-save adder, four iterations are conducted for each level, and parallel computing is conducted on the last 32 iterations; format conversion is conducted a CORDIC iteration result according to the characteristics of the function. The device comprises a pre-processor, a mixed mode CORDIC computing module and a standardization processing module. The method has the advantages that the method is simple in structure, the circumference coordinate CORDIC algorithm and the hyperbolic coordinate CORDIC algorithm can be executed on the same hardware platform, delay is low, a period is short, speed is high, and precision is high.

The invention relates to the field of digital marketing, in particular to a digital middle-platform system, a construction method and an application method, and the middle-platform system comprises aglobal member module, a full-contact marketing module, a full-channel transaction module, a full-link service module, a business middle-platform module, a data middle-platform module and a technical platform module. A full-link digital marketing function can be realized through the global member module, the full-contact marketing module, the full-channel transaction module and the full-link service module; and the business middle-platform module and the data middle-platform module provide support for realization of the full-link digital marketing function. And the technical platform module provides construction support and management for each module of the digital middle-platform system. The digital middle-platform system adopts a core design concept of agility, decoupling sharing and highreusability based on a large middle-platform and a small front end, realizes a full-link digital marketing business closed loop, leads enterprise digital innovation, and is high in integrity and goodin planarity.

The invention belongs to the technical field of image exposure, and relates to an automatic exposure control method. The automatic exposure control method comprises the steps of S1, carrying out imagecollected by using binocular vision and an active structural light source projection method, and obtaining the last image collected each time; S2, obtaining the average brightness of the image by using a dynamic partitioning and dynamic weight coefficient adjusting method; S3, judging whether the average brightness of the image is within a target brightness range or not, if so, not carrying out exposure compensation and terminating the automatic exposure control, otherwise, entering the step S4; and S4, obtaining the difference between the average brightness of the image and the target brightness, selecting the corresponding step size for exposure compensation according to the difference and returning back to the step S1. The automatic exposure control method soles the problems of great noise signal, insufficient or excessive exposure of a target area due to excessive contrast between the target object and the background, excessively long adjustment time and excessively great difference between the image brightness and the target brightness in the prior art.

The invention discloses an ISOA-LSSVM-based subway air-conditioning system energy consumption prediction method. The method includes the following steps: acquiring training data, standardizing the training data, using an improved population search algorithm to conduct parameter optimization on a least squares support vector machine, and establishing a prediction model; acquiring real-time measurement data and standardizing the real-time measurement data, inputting the standardized real-time measurement data to the prediction model and performing prediction, and finally performing reverse standardization and outputting a predicted energy consumption value. According to the invention, the method can predict ISOA-LSSVM-based subway air-conditioning system energy consumption; the improved population search algorithm uses a Gaussian membership function to represent a fuzzy variant of step size in search, reduces the times of iteration, and increases prediction precision of the model; the preliminary direction is obtained by comparing individual optimal fitness value and the fitness value of a current individual, and the obtained preliminary direction can better represent the preliminary action of the current individual and at the same time the iteration speed is increased.

The invention discloses a machine learning system. The system comprises clients used to collect to-be-recovered data and a cloud-end server connected with the client via network. Each client comprises a first processor provided with an improved intelligent algorithm; and the cloud end server comprises a second processor provided with a BP neural network model, the second processor processes the to-be-recovered data via the BP neural network model and optimizes initial parameters value of the BP neural network via the improved intelligent algorithm, and the initial parameter values include an initial weight and an initial threshold. The machine learning system is higher in learning speed and learning precision.

The invention discloses a particle swarm algorithm-based virtual-machine deployment method under a cloud environment, and belongs to the field of resource scheduling under cloud computing environments. For solving problems of only carrying out optimization of a single objective and lacking consideration of multiple objectives; the invention provides the particle swarm algorithm-based virtual-machine deployment method under a cloud environment. According to the method, deployment mapping between virtual machines and physical machines is established at an IaaS (Infrastructure as a Service) layer. The method of deployment mapping includes: a user-oriented virtual-machine deployment method, which includes receiving an application of a user for a virtual machine, and deploying the same to an objective physical-host on the basis of an improved multi-objective-optimization particle swarm optimization algorithm of congestion degree judgment; and a platform-oriented virtual-machine dynamic-management method, which includes deploying a virtual machine to an objective physical-host, and then judging whether status of the objective physical-host is above or below a normal threshold value, anddetermine a mapping relationship of the objective physical-host and the virtual host on the basis of an improved multi-objective-optimization particle swarm algorithm of sharing degree judgment. The virtual-machine deployment method is used for deploying the virtual machines on the objective physical-hosts.

A decoding device having a turbo decoder and an RS decoder concatenated serially and a method of decoding performed by the same. A turbo decoder decodes received data of a channel and an RS decoder RS decodes the turbo decoded data. A controller controls the turbo decoder to iteratively turbo-decode the data according to a number of iterations determined by a stored iteration number and to cease the turbo decoding if an error correction completion signal is received from the RS decoder. The controller decreases the iteration number for a next frame of the data if the completion signal is received within the predetermined number of iterations and increases the iteration number for a next frame if the completion signal is not received within the predetermined number of iterations. The iteration number is changeable within maximum and minimum limits and may exceed the maximum limit in special cases.

The invention discloses a motor imagery EEG pattern recognition method based on the time-frequency parameter optimization of an artificial bee colony. The method comprises the steps of conducting the leads selection based on the linear decision rule, selecting time-domain and frequency-domain optimal parameters based on the artificial bee colony algorithm, extracting features based on the common spacial pattern algorithm, and finally classifying features based on the linear discriminant analysis algorithm. The result of the method shows that, a lead channel of larger inter-class distinction degree can be effectively selected based on the lead selection algorithm. At the same time, based on the time-frequency parameter optimization algorithm of the artificial bee colony, a time window and a frequency band of larger inter-class distinction degree can be automatically selected, so that a better classification effect is obtained. The method is capable of effectively recognizing different motor imagery modes. Compared with the traditional parameter manual selection method and the frequency-domain parameter automatic selection algorithm, global optimal parameters can be automatically searched in both time domain and frequency domain at the same time based on the above method. Therefore, the feature extraction and feature classification effect for motor imagery EEG signals is improved.

The invention discloses a rapid probabilistic load flow calculation method considering the static power frequency characteristics of an electric power system. The rapid probabilistic load flow calculation method comprises the steps that electric power system parameters needed by conventional load flow calculation are extracted from the electric power system and are initialized; a frequency variable to be solved is added to a conventional load flow calculation model, an improved rapid decoupling load flow module is established, and the normal state of the node voltage, the system frequency and the branch power of the variable to be solved are worked out; cumulants of each order of the node voltage, the system frequency and the branch power are worked out; through Gram-Charlier series expansion, the cumulative probability distribution function and the probability density function of the node voltage, the system frequency and the branch power of the variable to be solved are worked out. According to the rapid probabilistic load flow calculation method, the influence of uncertain factors on the system frequency in the electric power system and the distribution characteristics of the system frequency are considered in the process of probabilistic load flow analysis, the calculation speed is high, and a complete comprehensive assessment can be provided for the safe and economical operation analysis and the stability analysis of the electric power system.

The present invention provides an automated design method and platform for intelligent hardware system development. The method comprises: creating various types of module resource libraries; performing demand analysis on a newly-developed product; forming a system parameter configuration table according to various requirements of the newly-developed product; importing the parameter configuration table into an automated design platform; enabling the platform to call a module resource library to perform scheme optimization; and according to the parameter configuration table, generating, by configuration, various types of documents, a circuit principle diagram, a circuit PCB layout, a hardware drive program, a mobile terminal application and a cloud server program, so as to perform debug detection. The platform is an operating platform of the method, and the platform comprises a plurality of application functional modules and module resource libraries. Powerful module resource libraries are built in the platform, and non-professional personnel can implement type selection of various types of hardware components by means of product functions and performance parameters; various types of software sub-modules and hardware sub-modules have a strong called property and high reusability, and the hardware is short in development cycle and high in development success rate; and excessive professional personnel are not required, and the research and development costs are low.

The invention discloses a method for preparing a three-dimensional microfluidic chip, and belongs to the technical field of microfluidics. Water-soluble materials are selected, and a hollow model of internal three-dimensional passageways of the microfluidic chip is processed based on a 3D printing technology; the coarse degree of the outer surface of the hollow model is reduced through dipping andrapid drying process; secondary pouring is conducted on polydimethylsiloxane, after curing is conducted, reserved edge banding of the outlet and inlet of the hollow structure is cut, and the water-soluble materials are dissolved by using hot water bathing to obtain the microfluidic chip with a three-dimensional passageway structure and a smooth inner wall surface. The method has the advantages oflow implementing difficulty, simple procedures, low cost and the like, the prepared three-dimensional microfluidic chip has a highly complicated three-dimensional structure, the coarse degree of theinner surface of the three-dimensional microfluidic chip is low, and the visibility degree is high.

The invention discloses a method for estimating an equivalent quality factor and a method for estimating a stratum quality factor by using the method for estimating the equivalent quality factor. A binary algorithm-based center-of-mass frequency shift quality factor value Q estimating method is provided on the basis of an attenuation formula of a quality factor value Q and the definition of a center-of-mass frequency and on the basis of determining the changing monotonicity of the center-of-mass frequency along with the quality factor value Q. According to the method for estimating the equivalent quality factor and the method for estimating the stratum quality factor according to the method for estimating the equivalent quality factor, the assumption that the seismic source frequency spectrum in the calculation of the quality factor value Q is a Gaussian spectrum in the conventional center-of-mass frequency shift method can be eliminated; the application range of the center-of-mass frequency shift method is expanded; the calculation precision of the quality factor value Q is improved.

The invention discloses a genetic algorithm-based RV speed reducer tooth profile optimizing and modifying method. The tooth profile equation of a cycloidal gear after modifying and the contact range during actual work are determined according to the traditional three modifying methods of the cycloidal gear; the corresponding target function and the constraint condition are determined by the engaged nominal clearance of the tooth profile after modifying; the optimal modifying quantity is compensated into the tooth profile of the cycloidal gear of an RV speed reducer, so that the genetic algorithm optimization model of the tooth profile of the cycloidal gear is obtained. The optimal equidistant modifying quantity and the shift modifying quantity serve as target variables, the minimal tooth profile clearance after two kinds of optimal modifying serves as the target function, the engaged clearance between the node of the cycloidal gear and the engaged point serves as the constraint condition during optimization, the mixed penalty function of two-point heterodyning is introduced, so that the iteration speed of the optimization algorithm is increased, and finally the optimal equidistantand shift modifying quantity is obtained. By the method, optimal computation parameters can be determined rapidly and effectively, and the computation efficiency and the optimization precision are greatly improved.

A method for multidimensional scaling (MDS) of a data set comprising a plurality of data elements is provided, wherein each data element is identified by its coordinates, the method comprising the steps of: (i) applying an iterative optimization technique, such as SMACOF, a predetermined amount of times on a coordinates vector, said coordinates vector representing the coordinates of a plurality of said data elements, and obtaining a modified coordinates vector; (ii) applying a vector extrapolation technique, such as Minimal Polynomial Extrapolation (MPE) or reduced Rank Extrapolation (RRE) on said modified coordinates vector obtaining a further modified coordinates vector; and (iii) repeating steps (i) and (ii) until one or more predefined conditions are met.

The embodiment of the invention discloses an AB test method and device. The method comprises a step of determining a test strategy for each test layer in a plurality of test layers, wherein the plurality of test layers are used for determining a target test strategy of each request flow in each test layer, each test layer in the plurality of test layers comprises a reference test and a test to betested, and the test strategy determination step comprises the following steps: obtaining the request flow of at least one user in response to a trigger request of at least one user in a target application; determining a target test corresponding to the i request flow in the request flow of at least one user in the test layer, and distributing the i request flow to the target test; and determining the test strategy in the target test as the corresponding target test strategy of the ith request flow in the test layer, i = 1,... N, and N being the number of the request flows of atleast one user. According to the embodiment of the invention, the problem of long optimization iteration period of a product caused by the fact that multiple tests cannot be carried out by a test method in the prior art is solved.

The invention relates to a multi-terminal shared bottom framework generation and application method based on a Lua script technology. Compared with the prior art, the method overcomes the defect thata plurality of terminals cannot share a bottom framework. The method includes the following steps that a Lua engine is initialized; a UI layer is created; a Lua layer is created; the UI layer is correlated with the Lua layer; a database layer is correlated with the Lua layer. Different development modes of multiple platforms are fused on a bottom layer, correspondingly the attention of each platform is only focused on implementation of the UI layer, and therefore there is no need to pay attention to formation of the bottom framework. Thus, the development time of each platform is greatly shortened, the complexity of each platform is greatly reduced, and the iterative speed is increased.

An imaging method providing atmospheric turbulence compensation includes capturing a subject image pair. Data corresponding to the captured subject image pair is loaded into memory associated with a parallel processing device. A 2-D Fast Fourier Transform is performed on the stored image data and the transformed image data is stored. An optical transfer function is developed for the transformed image data. The optical transfer function is inverted and applied to the transformed image data to generate corrected image spectrum data that is compensated for atmospheric aberrations. An Inverse Fast Fourier Transform is applied to the corrected image spectrum data to produce corrected image data. The corrected image data is stored.

The invention provides a real-time video code rate self-adaptation method based on mobile edge calculation. The method comprises the following steps: S1, acquiring parameters such as video playing, network connection and equipment performance of a user in real time; S2, selecting a task scheduling strategy with the optimal performance for the user according to the parameters in S1; S3, automatically switching working modes according to the selected task scheduling strategy; S4, calculating the QoE of the current scheduling strategy according to the parameters such as video playing condition inthe next iteration period, the equipment energy consumption and the like, and feeding back the QoE to the edge server; S5, by the edge server, updating a scheduling strategy state transition matrix according to the feedback QoE; and S6, repeating the above steps, and continuously performing iterating and updating, so that the performance of the final edge scheduling strategy tends to be optimal.According to the invention, on the basis of relatively low cost, relatively high energy consumption can be reduced for mobile multimedia equipment, and continuous growth and optimization of a complexvideo code rate adaptive algorithm learning system are helped.

The invention provides a voice scene updating method, device, terminal, server and system. The updating method comprises the following steps: sending an updating request to a cloud, wherein the updating request is used for enabling the cloud to acquire version information of the vehicle-end voice scene data and version information of at least one piece of cloud voice scene data in real time so asto determine whether updatable voice scene data exists or not, and determining target voice scene data under the condition that the updatable voice scene data exists, wherein the at least one piece ofcloud voice scene data comprises new-version cloud voice scene data obtained by updating the original-version cloud voice scene data based on non-feedback voice information, and the non-feedback voice information is voice which cannot be fed back based on the vehicle-end voice scene data; receiving target voice scene data returned by the cloud; and updating the vehicle end voice scene data into the target voice scene data. According to the technical scheme provided by the invention, the iteration speed of the voice scene data and the updating efficiency can be improved.

The invention belongs to the field of short text auditing, particularly relates to a short text auditing method and device fusing variant word recognition, and aims to solve the problem of how to fusea variant word recognition technology into a harmful text auditing task and realize automatic updating of a model. The method comprises the steps of constructing a configuration word bank, obtainingto-be-audited text data based on a social media platform, screening the to-be-audited text data to obtain suspicious text data, removing meaningless information, and calculating a text feature vectorand a statistical feature vector; and performing feature fusion on the text feature vectors and the statistical feature vectors, obtaining harmful texts through a trained harmful text classification model based on a support vector machine, obtaining sensitive words of the harmful texts by utilizing a preset keyword extraction algorithm, and writing the sensitive words into a configuration word bank. According to the method, the variant word recognition technology is fused into text feature and statistical feature calculation to carry out harmful text auditing tasks, automatic model updating isachieved, and the text auditing accuracy and updating speed are improved.

The invention provides an ICD (Interface Control Document) integrated management tool. The ICD integrated management tool (namely an internal integrated design tool) consists of a definition design tool, a version maintenance tool, an interface output conversion tool and a system authority review tool, and is connected with a database management tool through a local area network, wherein the definition design tool is used for setting input and output of each module of a project management system and a bus management system in a grading manner in a visual design mode and an intuitive bytes and bit fields parallel design mode; the version maintenance tool forms a whole set of finished version specifications for ICD maintenance by virtue of different version repositories generated in an ICD; the interface output conversion tool is used for converting the ICD into an XML (Extensive Makeup Language) document describing the ICD by virtue of an XML document output system, and provides a whole set of complete conversion interface; and the system authority review tool guarantees the hierarchical authority and the confidential level of the ICD. By virtue of the ICD integrated management tool, the design of the ICD can be simpler, the management of the ICD can be more systemized, and the maintenance of the ICD can be more traceable.

The invention provides a binocular video depth map calculating method. The method includes the following steps of carrying out edge detection on two pictures, carrying out picture segmentation on the two pictures, carrying out classification on detected edges by using of picture segmentation results, and carrying out depth calculation through improvement of a BP algorithm. According to the method, edge detection results are obtained through improving of a canny operator, then relation of the edges and the segmentation results is analyzed, the BP algorithm is improved, single transmission is carried out in different areas, and a message transmission mechanism is changed; single processing is carried out on a shielded part so as to enable the shielded part not to influence transmission of correct messages, the iteration speed is increased, and precision is improved.

The invention discloses a multicore system task scheduling method based on particle swarm optimization, comprising the steps of constructing a task scheduling model to generate an DAG (directed acyclic graph), initializing basic parameters and pheromones, updating the pheromones to calculate inter-task communication time and fitness value, and generating a global optimal solution. Ant colony optimization and particle swarm optimization are combined by effectively initializing the pheromones, and a heuristic intelligent optimization algorithm is used to efficiently schedule tasks for a multicore system. The method of the invention has the advantages that task scheduling time is shortened, communications cost is lowered, and defects of a unilateral algorithm are made up.

The invention designs a diagonal recurrent neural network (DRNN) control strategy (Q-DRNN) based on a Q learning algorithm. The Q-DRNN organically combines the strong search capability of Q learning with the advantages of the built-in recursive ring structure, the dynamic mapping capability, the adaptive time-varying characteristic and the like of the DRNN, and is used for improving the working stability of a brushless direct current motor (BLDCM). In the Q-DRNN, the DRNN iterates an output variable through a unique recursive ring in a hidden layer, and optimizes the key weight of the output variable so as to accelerate the iteration speed; and meanwhile, the improved Q learning is introduced to correct the weight term factor of the DRNN, so that the DRNN has self-learning and online correction capabilities, the anti-interference capability and robustness of the system are enhanced, and the brushless direct current motor achieves a better control effect.