898results about How to "Reduce time overhead" patented technology

The invention discloses a topic detecting device and a topic detecting method based on distributed multistage cluster. The topic detecting device mainly comprises a news acquiring module, a news classifying module, a topic detecting module, a topic integrating module and a topic displaying module. The topic detecting method is characterized by comprising the steps of A, acquiring news; B, classifying the newly acquired news; C, performing multistage cluster for various channels concurrently; and D, computing hotness of all topics, and screening hot topics of a total system and hot topics in each channel. By the topic detecting device and the topic detecting method, a sharp contradiction between a detecting effect and time cost in topic detection under the condition of quick updating of a large number of files in an internet environment can be solved.

The invention provides a CNN-based low-precision training and 8-bit integer quantization reasoning method. The method mainly comprises the steps of carryin gout low-precision model training; Performing model training by using a 16-bit floating point type low-precision fixed point algorithm to obtain a model for target detection; Quantifying the weight; Proposing an 8-bit integer quantization scheme, and quantizing the weight parameters of the convolutional neural network from 16-bit floating point type to 8-bit integer according to layers; carrying out 8-bit integer quantitative reasoning; quantizing the activation value into 8-bit integer data, i.e., each layer of the CNN accepts an int8 type quantization input and generates an int8 quantization output. According to the invention, a 16-bit floating point type low-precision fixed point algorithm is used to train a model to obtain a weight; Compared with a 32-bit floating point type algorithm, the method has the advantages that the 8-bit integer quantization reasoning is directly carried out on the weight obtained by training the model, the reasoning process of the convolutional layer is optimized, and the precision loss caused by the low-bit fixed point quantization reasoning is effectively reduced.

The invention discloses an image-based vehicle anti-collision early warning method. An image sensor is utilized for collecting the images of a rod ahead; after a vehicle is detected successfully and tracked, according to the vehicle width variation in interframe images and combining with the influences of the relative vehicle speed and relative acceleration, the time to collision (TTC) of a target vehicle and the vehicle is calculated, the dangerous level is analyzed, and an alarm is given to a driver, and meanwhile, the relative driving state of the target vehicle and the current vehicle also can be determined. The method provided by the invention is applied to the technical field of vehicle active safety, can effective prevent rear-end collision with the vehicle ahead, also is not influenced by rugged topography, and has high safety performance.

The invention discloses an active learning based data automatic marking method, and belongs to the field of active learning. The data automatic marking method comprises the steps of 101, processing marked data and unmarked data; 102, classifying the unmarked data by using a plurality of different classifiers; 103, selecting data with low difference entropy; 104, performing manual marking on the data with low difference entropy; and 105, performing self-checking on a manual marking result. The invention provides a data automatic marking system with a self-checking function through combining an active learning method in allusion a problem of how to ensure the accuracy of manually marked data as far as possible while reducing the volume of the manually marked data, thereby achieving the purposes of reducing the workload and improving the accuracy of the manually marked data.

The invention discloses a random sampling consistency-based characteristic line detection method for a three-dimensional point cloud. The method comprises the following steps of: detecting a plurality of planes in the point cloud based on random sampling consistency; and applying the random sampling consistency to boundary points, which are taken as candidate points, of each plane parameterization area so as to obtain a final characteristic line. Therefore, the influences of noise, exterior points and data loss are reduced effectively, the detection robustness of the characteristic line is enhanced greatly and time efficiency is improved.

A technique for monitoring an execution of a program is provided. At least one relational query is constructed and compiled into a program instrumentation. The program instrumentation is incorporated into the program and one or more query results are output from the program as the program executes.

Streaming multimedia content from a content server to client, user multimedia devices (UMDs) connected to the content server through a network includes the steps of monitoring storage states, of UMDs connected to the network, for preset multimedia content, and automatically transmitting a starting block of the content to any UMD not yet storing the block. Tasks preparatory to streaming are advantageously off-loaded to the client side and distributed among UMDs, to realize reduced bandwidth and delay and to conserve on storage capacity of a UMD.

The invention relates to an on-line large power grid recovery assistant decision-making system. The system comprises two work modes, namely a real-time assistant mode and a research and analysis mode, a fracture surface operation analysis module, a calculation configuration module, a calculation function module and a calculation management module, wherein the fracture surface operation analysis module is used for fracture surface reading and fracture surface adjustment, the calculation configuration module is used for setting calculation parameters and selecting a black-start power supply and a started power supply, the calculation function module is used for system partitioning, load flow calculation, electromagnetic transient calculation, electromechanical transient calculation and path search and assessment, and the calculation management module is used for displaying system partitioning schemes, voltage waveforms, ranking of available paths and a black-start generating scheme. When the on-line large power grid recovery assistant decision-making system is used for developing an on-line electric power system recovery assistant decision-making supporting system, existing data resources in a dispatching center can be sufficiently utilized, the advantages of the method of pre-arranged scheme recovery and intelligent dispatching are combined, a model does not need to be rebuilt, the construction time and cost of the system are reduced, the real-time performance and the accuracy of suggestions to decision-making are improved, and a powerful support is provided for rapid and reliable recovery of a power grid after power outages.

The invention discloses an HLA-based distributed simulation support platform and an implementation method thereof, which are designed to solve the defects of the conventional HLA distributed simulation system in terms of operating performance and communication flow. The HLA-based distributed simulation support platform comprises a high-performance computing cluster compatible RTI (Run-Time Infrastructure) system unit, a reflective memory network and a federate deployment module, and provides HLA interface specification-based service for a simulation model. The implementation method for the HLA-based distributed simulation support platform is as follows: an interface function library provides service requests for federates via an RTI ambassador process, and federate processes submit the requests to an RTI via an RTI ambassador; the federates are interconnected and interoperated in a local area network, and a local RTI server centrally carries out management and communication. The invention achieves the state that the whole simulation federates can coordinatively advance, and increases the efficiency of distributed simulation; and the reflective memory network accelerates data synchronization.

Methods and apparatus for efficient transmission of data by half-duplex transceivers in satellite communication systems are provided. Time reference for the return link is skewed or time-lagged relative to the time reference for the forward link to reduce the amount of guard time required to separate return link transmission from forward link reception by the half-duplex transceiver of a user terminal. The guard time is determined based on a maximum differential round-trip propagation delay and transition times of the half-duplexer transceiver to switch between transmit and receive modes. In a satellite communication system in which a large number of active user terminals are present in a beam coverage, random time offsets are applied to spread approximately equal traffic loads across the time offsets.

The invention discloses a dynamic operation scheduling system of a virtual machine. The system can be conveniently connected to various cloud platforms to supply unified resource management for the various popular cloud platforms. The system can dynamically allocate the operation pretreatment strategy, placement strategy and integration strategy of the virtual machine, can be realized in a plug-in mode and has high dynamic characteristics and good extensibility. In comparison with the prior art, the system provided by the invention not only has the advantages of dynamic allocation and suitability to various cloud platforms, but also can schedule and execute the preparation operation before turning on the virtual machine as well as the cleanup operation after turning off the virtual machine in a task form, thereby improving the deployment efficiency of the virtual machine and improving the resource utilization rate. In addition, the system has good extensibility, and supports heterogeneous virtual machine platform, thereby improving the transparence of resource utilization.

The invention discloses a virtual machine mirror image generating method and a device based on software preinstallation in cloud environment. The virtual machine mirror image generating method and the device based on the software preinstallation in the cloud environment comprise that softwares which are uploaded by users are received and dependency of the uploaded softwares are checked, operating system configuration information which is submitted by the users is received and analyzed to obtain analysis results of the operating system configuration information. Basic environment which is necessary to install an operating system based on the software preinstallation technology is generated according to the analysis results of the operating system configuration information. A virtual machine start is started by using a virtual machine start configuration file in the generated environment and a monitoring progress is started at the same time. An operating system installation program is operated in the virtual machine and the operating system is installed in a non-interactive mode. Meanwhile, the monitoring progress judges whether the installation of the operating system is wrong or accomplished or not. When the installation of the operating system is wrong, the virtual machine is closed and temporary files in the basic environment are cleaned and internet protocol (IP) address is returned to an IP address pool. The virtual machine mirror image generating method and the device based on the software preinstallation in the cloud environment can solve defects of an existing manner, wherein the defects of the existing manner are insufficient in flexibility, low in speed, serious in source consumption and the like.

The invention provides a personal interactive data retrieval method and a personal interactive data retrieval system based on a tag technology. According to the method, a binary relation between interactive data and tags is stored in a database; by designing a method for calculating the total weight of the tags and the weight of each tag, a user interest model based on a weighted tree structure is constructed, so the tag preference characteristic of a user is effectively described; when the user adds the tags to resources, the corresponding tags are intelligently output to the user according to the user interest model; and meanwhile, an association relation between the resources is built according to the resource interactive operation characteristic of the user and co-occurrence of the tags, the resources are recommended and output to the user, and the browsing and retrieval efficiency of the user is improved. By the invention, user requirements on personalization and intelligentization of personal interactive data management can be met, and the interaction burdens of the user are effectively reduced.

The invention discloses a local sampling-based multi-geometrical characteristic point cloud data splitting method. The method includes the following steps: a three-dimensional rule grid is utilized to divide all acquired point cloud data into a plurality of first grid units; the process of extracting each geometrical characteristic includes the following steps: a sampling point is randomly chosen from the first current data set, the current first grid unit with the sampling point is determined, the sampling point and the other points in the current first grid unit are utilized to construct Alpha candidate geometrical models, an optimal model is determined from the Alpha candidate geometrical models, a consistent set of the optimal model is calculated in the first current data set, and according to multiple geometrical characteristics, all the point cloud data are split into a plurality of subsets. From local sampling, the method constructs candidate geometrical models from a first grid unit, and determines an optimal model from the candidate geometrical models, so that a geometrical characteristic is extracted, thus the problem of fitting models not existing in reality is prevented; and the method is more efficient.

The invention relates to a real time multiplexed synchronous high speed transmission serial bus protocol, wherein transmission signals of a series bus in the protocol comprise three groups of signals of control signals, clock signals and data signals, any node can be appointed in advance as an only main node on the bus, the control signals are sent by the main node to control each node to send data in turn, the sending order and the data length of each node are preset, all data on the bus can be received through the clock signals and the data signals, the control signals of the main node comprise three states: (a) idle states, a synchronous series bus under the states is controlled by the main node, (b) node switching states, the states represent that data transmission of a preceding nodeis finished, the transmission of a next node is about to start, the node which obtains the bus control right discards the bus control right under the states, a next bus owner obtains the bus control rights and starts outputting the bus signals, (c) data transmission states, the states represent that the node which obtains the bus control right is outputting signals to the bus.

The invention relates to the technical field of data processing, and discloses a distributed file processing method and device. The method comprises the following steps: a metadata server receives a file reading and writing request sent by a terminal, wherein the file reading and writing request is sent according to a shared directory of the metadata server mounted in the terminal; and the metadata server sends the file reading and writing request to a storage cluster of a file system according to the file system mounted in the shared directory, wherein the metadata server and the storage cluster are located in a same file system cluster. The method and device provided by the invention aims to solve the problem in the prior art that the terminal with a different operating system cannot access the file system, or a process in which the terminal accesses the file system is complicated and time-consuming.

The present invention relates to a system and method of data management for controlling a plurality of peripheral devices. In an example embodiment, a data management device for managing data communications between a data control device and the plurality of peripheral devices is added between them, and the data management device controls the plurality of peripheral devices by grouping them according to their characteristics, whereby various protocols can be integrated into minimized optimal protocols by the grouping, and the data control device can reduce technical and temporal overheads due to detection and control of the plurality of peripheral devices.

The invention discloses an LTE broadband communication system calculation resource dispatcher based on a GPP and a dispatching method thereof. The calculation resource dispatcher comprises three parts which are an LTE assembly state information storer, an LTE calculation resource dispatching center and an LTE calculation resource pool. According to the calculation resource dispatcher, in order to meet the strict requirements of an LTE broadband communication system for time delay, in the process of adoption of assembly line mode processing communication signals based on the GPP, the cloud calculation conception is utilized, base stations are centralized together to form a base station cluster, in the base station cluster, the GPP is used as a calculation resource, the GPP calculation resource are dispatched, a large number of calculation tasks are divided into reasonable granularity and then are distributed into assembly lines of various grades for processing so as to enable the assembly line mode processing communication signals to be capable of meeting the instantaneity requirement, and time safety surplus of all task operation is increased. In order to solve the problem that the overhead of Turbo decoding time is large, the invention further discloses the dispatching method for reasonably distributing Turbo decoding code blocks, so that the real-time performance of the LTE broadband communication system is further increased and improved.

A method and data structure for transmitting data in a networked system having a shared transmission medium that includes generating a plurality of virtual packets each containing information for delivery to at least one selected terminal in the networked system via the shared medium and generating a map field having control information for controlling the delivery of the virtual packets. The method and data structure also include forming a physical packet for transmission across the shared transmission medium, the physical packet including a control field and the plurality of virtual packets such that overhead associated with the control field and medium access is amortized over the plurality of virtual packets.

The invention discloses an optical receiver which comprises a receiving unit, a converting unit, a detecting unit and a burst clock data recovering circuit, wherein the receiving unit is used for receiving optical signals which at least comprise optical signals having two different transmission rates; the converting unit is used for converting the received optical signals to voltage signals; the detecting unit is used for detecting the transmission rates of the voltage signals and outputting control signals according to the detecting result; and the burst clock data recovering circuit is usedfor recovering the clock and the data for the voltage signals according to the control signals. Correspondingly, the invention also discloses a method for receiving optical signals. By the invention,not only time multiplexing signals with two rates of GEPON and 10G EPON can be received, but also fast synchronization and time saving can be realized.

A read command having designated a bank, can be inputted from outside. A read command having designated a bank can be inputted from outside while an operation for reading from a memory array to a data buffer is being performed at the bank. Further, a read command having designated a bank is inputted from outside, and a buffer read command having designated a bank is inputted from outside while an operation for reading from a memory array to a data buffer is being performed at the bank, whereby reading from the data buffer of the bank to the outside is enabled.