32results about How to "Optimize execution time" patented technology

The invention belongs to the technical fields of cloud computing and data mining, and particularly discloses a massive high-dimension data clustering method for a MapReduce platform. In the method, each dimension of raw data is split, and clustering is performed by utilizing small split non-null grids instead of points in the raw data so as to reduce a data scale. The clustering is realized by utilizing an open source of MapReduce, so that the whole clustering process can be finished in parallel on a distributed cluster, and the limitations of a single-machine algorithm to storage and computation are broken. In the clustering process, the thought of a K-mediods algorithm is adopted, and a highly-efficient Euclidean distance computation method is put forward. The method is applied to the processing of massive high-dimension data. A user can perform manual regulation on the algorithm according to the computational capability of the cluster, the expected time of the algorithm and requirements on clustering accuracy. The needs of different users are satisfied.

The invention relates to the technical field of cloud computing, in particular to a workflow scheduling and data distribution method under a cloud environment. The method comprises the following stepsthat current workflow of a cloud platform is obtained; the height of each task in the workflow is calculated according to a scheduling model, and the tasks are arranged in an ascending order according to the heights; a particle swarm optimization algorithm is used, the execution time and execution cost of the workflow serve as an optimization objective, and a globally optimal particle code is obtained; the globally optimal particle code is decoded into a workflow scheduling and data distribution scheme and output. Accordingly, data storage is considered, the time of across-node transmission is shortened, and the task execution efficiency and execution cost of the workflow in the cloud platform are optimized.

The disclosure relates in particular to the processing of commands targeting at least one element of a cluster including a plurality of elements, the at least one element having a link of dependency according to the at least one command with at least one other element. After having identified the at least one element and at least one dependency rule from the at least one command, a dependency graph is generated from the at least one identified element, by applying the at least one identified dependency rule, the dependency graph including peaks representing at least the element and the at least one other element, an action linked with the at least one command being associated with the peaks of the dependency graph. A sequence of instructions is then generated from the dependency graph.

A power outage plan optimization decision method aiming at reducing power outage time and quantity comprises the following steps: reading the topology information of a distribution network, and establishing the contact relationship between distribution network components in a region to be evaluated to form a contact matrix; reading the information of each load point in the distribution network; reading power outage event information; for each type of power outage events, reading the information of maintenance teams in the region to be evaluated (including the number BN of the maintenance teams in the region to be evaluated, the ability of the maintenance teams and the daily workload of the maintenance teams), and making clear the power outage strategy in the region to be evaluated; optimizing the power outage period; optimizing the power outage order; and displaying the result. The method takes the user's power outage time, the user satisfaction and the team's ability into full consideration, and adopts simultaneous maintenance. For different power outage events, the execution time of each power outage time is optimized based on the daily load curve of an affected region. Thus, the power outage time and power outage quantity of the system are reduced effectively, reliable power supply of the system is ensured, and the customer satisfaction is improved.

A method and apparatus for the processing of at least one command concerning at least one component of a cluster is disclosed. The cluster comprises several components, the at least one component having a dependency link, according to the at least one command, with at least one other component, is disclosed.

According to an embodiment, there is provided a method of generating a data structure stored in computer memory for processing a search query in a network of interconnected nodes, wherein the method comprises selecting landmark nodes by the following steps and storing the selected landmark nodes in the data structure: sampling from the network nodes a first sample of vertex pairs, computing the shortest path for each vertex pair, each shortest path comprising a set of vertices between each vertex in the vertex pair; identifying a first landmark node which occurs in more of the shortest paths more often than any other vertex; removing from the network vertices shortest paths including the first landmark node and identifying a second landmark node which occurs in more of the remaining shortest paths than any other remaining vertex.

A vehicle includes a first motor generator that generates electric power using power of an engine, a battery, and a second motor generator that is connected to drive wheels and is driven by electric power supplied from the battery and the first motor generator. An ECU of the vehicle drives the first motor generator as an electric motor with a regenerative electric power, executes a braking control, in which a load of the first motor generator functions as the engine, when an SOC is equal to or more than a threshold value of a waste electric-power start, and changes the threshold of the waste electric-power start based on a running condition of the hybrid vehicle.

The invention provides an acceleration optimization method and system for parallel PME of molecular dynamics simulation software, and the method comprises the steps: a calculation task is divided intoa plurality of parts, the calculation task is distributed to all slave cores of a super calculation platform, and the calculation is carried out through the slave cores; in the fast Fourier transformin the calculation process, data replication of matrix transposition is completed by utilizing data partitioning and a slave core; and the master core obtains the calculation results of the slave cores in a DMA mode, merges and vectorizes the calculation results, and the master core uses an RDMA technology to replace a traditional MPI technology to obtain data of other master cores. According tothe method, the molecular dynamics simulation software can be smoothly loaded on the super computing platform, and meanwhile, the computing performance can be improved.

An electronic device is provided such that a user can experience a quick launch of an application therein. The electronic device includes a housing, a display, an input unit, a processor, a non-volatile memory to store an application program, and a volatile memory to store instructions that allow the processor to load a first part of the application program in the volatile memory based on a firstchange of state of the electronic device, to load a second part of the application program in the volatile memory based on a second change of state of the electronic device and to display an image ortext generated by the loaded first or second part. Since at least part of the application is preloaded before the second input is generated, only the remainder of the application has to be loaded in order to execute the application after the second input is generated.

The embodiment of the invention provides an image processing method and device. The method comprises the following steps: converting a foreground image corresponding to an original image into a compressed image; estimating a first extraction time required for carrying out depth feature extraction by adopting the original image and a second extraction time required for carrying out depth feature extraction by adopting the compressed image; and according to the first extraction time and the second extraction time, determining to adopt the original image to carry out depth feature extraction or adopting the compressed image to carry out depth feature extraction. By adopting the embodiment of the invention, the calculated amount of the image depth feature extraction process can be effectivelyreduced, so that the execution time of video object detection is improved.

A search query to provide a search result may be received, which identifies source and target nodes and an application for generating the search result. The application accesses a data structure holding landmark nodes, which store a shortest path tree in the form of a set of parent links. Each parent link can identify an adjacent vertex node in a shortest path between each node in the data structure and the landmark node. The location of the source node and the target node in the shortest path trees may be identified to the landmark node. For each landmark node, using the identified locations of the target node and source node, a measure of distance between the source node and the target may be generated. The landmark node with the shortest distance may be determined. A search result related to the shortest path tree of that landmark node may be provided.

The invention discloses a method for a broadcast bearing layer message filtering strategy based on SIG mesh. The method is based on a Bluetooth SIG mesh mesh network. The mesh mesh network comprises abearer layer, a network layer, a transport layer and an access layer, and is characterized in that the bearer layer is arranged on the mesh network; the bearer layer is used for transmitting data between mesh nodes and transmitting the data to the network layer; and the mesh node receives the broadcast message at the bearer layer, and the broadcast message is compared with a preset queue, so thatthe broadcast message in the mesh mesh network is filtered, and the RAM consumption and the CPU execution time are reduced. According to the invention, the RAM, the CPU execution time and the BLE scanning time consumed by the interference of the air non-mesh broadcast packet are optimized, and the RAM consumption and the CPU execution time caused by repeated data packets are optimized.

This engine system is provided with: an engine; an injector; a super charger (including a compressor); an electronic throttle device provided in an air intake passage, the compressor being provided in the air intake passage upstream of the electronic throttle device; an evaporated fuel treatment apparatus (including a canister, a purge passage, and a purge valve), an outlet of the purge passage being connected to the air intake passage upstream of the compressor; and an electronic control unit (ECU). The ECU controls the purge valve in order to perform a purge cut of the vapor from the purge passage toward the air intake passage when determining that the engine has started to decelerate, and thereafter controls the injector in order to perform fuel cut to the engine.

According to an information recording / reproducing method by which, at recording or reproduction of data, the data are temporarily stored in a memory to perform the recording or reproduction, when recording or reproduction of the data stored in the memory is performed with dividing data, recording or reproduction of subsequent data is performed without dividing the data, and when recording or reproduction of the data stored in the memory is performed without dividing data, recording or reproduction of subsequent data is performed with divided the data. Accordingly, in a disc device that records or reproduces video on / from a disc, when data stored in a memory is recorded or reproduced without dividing the data and thereafter subsequent recording or reproduction is performed, a rotational delay would not occur, and the increase in the recording / reproduction execution time resulting from the rotational delay is avoided, whereby the data recording / reproduction execution time can be optimized.

A vehicle includes a first motor generator that generates electric power using power of an engine, a battery, and a second motor generator that is connected to drive wheels and is driven by electric power supplied from the battery and the first motor generator. An ECU of the vehicle drives the first motor generator as an electric motor with a regenerative electric power, executes a braking control, in which a load of the first motor generator functions as the engine, when an SOC is equal to or more than a threshold value of a waste electric-power start, and changes the threshold of the waste electric-power start based on a running condition of the hybrid vehicle.

A method for validating a method for sterilizing an item, making it possible to validate the sterility assurance level achieved with this sterilization method. The method includes carrying out a first step of contaminating a container receiving the item with more than 105 living microorganism cells, then carrying out a first sterilization cycle with the chosen method, then opening the container in order to contaminate it again with more than 105 living microorganism cells, then carrying out a second sterilization cycle with the same method, and finally checking the sterility of the container after the first sterilization cycle and after the second sterilization cycle. The method is applicable in particular for products and devices intended for health use.

The invention belongs to the technical fields of cloud computing and data mining, and particularly discloses a massive high-dimension data clustering method for a MapReduce platform. In the method, each dimension of raw data is split, and clustering is performed by utilizing small split non-null grids instead of points in the raw data so as to reduce a data scale. The clustering is realized by utilizing an open source of MapReduce, so that the whole clustering process can be finished in parallel on a distributed cluster, and the limitations of a single-machine algorithm to storage and computation are broken. In the clustering process, the thought of a K-mediods algorithm is adopted, and a highly-efficient Euclidean distance computation method is put forward. The method is applied to the processing of massive high-dimension data. A user can perform manual regulation on the algorithm according to the computational capability of the cluster, the expected time of the algorithm and requirements on clustering accuracy. The needs of different users are satisfied.

An embodiment of the present application provides an image processing method and device thereof, wherein the method includes the following steps: converting the foreground image corresponding to the original image into a compressed image; estimating the first extraction time required for depth feature extraction using the original image and the second extraction time required for depth feature extraction using the compressed image; according to the first extraction time and the second extraction time, it is determined to use the original image for depth feature extraction or use the compressed image for depth feature extraction. By adopting the embodiment of the present application, the calculation amount of the image depth feature extraction process can be effectively reduced, thereby improving the execution time of video object detection.

A power outage plan optimization decision method aiming at reducing power outage time and quantity comprises the following steps: reading the topology information of a distribution network, and establishing the contact relationship between distribution network components in a region to be evaluated to form a contact matrix; reading the information of each load point in the distribution network; reading power outage event information; for each type of power outage events, reading the information of maintenance teams in the region to be evaluated (including the number BN of the maintenance teams in the region to be evaluated, the ability of the maintenance teams and the daily workload of the maintenance teams), and making clear the power outage strategy in the region to be evaluated; optimizing the power outage period; optimizing the power outage order; and displaying the result. The method takes the user's power outage time, the user satisfaction and the team's ability into full consideration, and adopts simultaneous maintenance. For different power outage events, the execution time of each power outage time is optimized based on the daily load curve of an affected region. Thus, the power outage time and power outage quantity of the system are reduced effectively, reliable power supply of the system is ensured, and the customer satisfaction is improved.

The invention discloses a task allocation method based on deep learning reasoning of a resource-limited robot, and aims to realize the task allocation of a multi-robot system to robot engineering tasks. The technical solution is to build a task allocation system, build a layer prediction model and store it on the robot; the task robot collects environmental data, the model interpretation module extracts the model layer type and related parameter configuration, and the resource consumption estimation sub-module estimates the level of each robot based on the layer prediction model. The delay of each layer of the deep learning model; the resource availability sub-module queries the resource status of the task allocation system; the decision-making module determines the optimal task allocation plan, and allocates engineering tasks to the robot nodes according to the optimal task allocation plan; The task allocation plan carries out the execution of engineering tasks. The invention can coordinate engineering task allocation between heterogeneous robots according to the task allocation system resource state, and optimize the execution time of the engineering task.

The invention provides a testing method and a testing device. The testing method includes: configuring execution sequence of a testing sample by using an automatic testing tool according to preset rules; using an industrial controlled device for controlling a Bluetooth system to be tested and an opposite terminal device to execute the testing sample according to the configured execution sequence of the testing sample, and obtaining the testing data; and comparing the testing data with preset testing standard data, and determining testing results according to the comparison results. According to the testing method and the testing device, problems that existing Bluetooth testing wastes manual power and is not high in efficiency are solved, and effects of high testing efficiency and low testing cost are achieved.

A method and apparatus for the processing of at least one command concerning at least one component of a cluster is disclosed. The cluster comprises several components, the at least one component having a dependency link, according to the at least one command, with at least one other component, is disclosed.

According to an embodiment, there is provided a method of generating a data structure stored in computer memory for processing a search query in a network of interconnected nodes, wherein the method comprises selecting landmark nodes by the following steps and storing the selected landmark nodes in the data structure: sampling from the network nodes a first sample of vertex pairs, computing the shortest path for each vertex pair, each shortest path comprising a set of vertices between each vertex in the vertex pair; identifying a first landmark node which occurs in more of the shortest paths more often than any other vertex; removing from the network vertices shortest paths including the first landmark node and identifying a second landmark node which occurs in more of the remaining shortest paths than any other remaining vertex.

The present invention relates to the technical field of cloud computing, in particular to a workflow scheduling and data distribution method in a cloud environment, comprising the following steps: obtaining the current workflow of the cloud platform; calculating the height of each task in the workflow according to the scheduling model, and Arrange the tasks in ascending order of height; use the particle swarm optimization algorithm to optimize the execution time and cost of the workflow to obtain the globally optimal particle code; decode the globally optimal particle code into a workflow scheduling and data allocation scheme and output. This method takes data storage into consideration, reduces the transmission time between nodes, and optimizes the task execution efficiency and execution cost of the workflow in the cloud platform.

An execution method when a CPU is applied to an artificial intelligence related program comprises the following steps that when TensorFlow is executed on electronic equipment, a corresponding artificial intelligence model is called according to program content; judging and extracting one or more sparse matrixes used by the artificial intelligence model in operation; performing a matrix simplification procedure on the one or more sparse matrixes; performing instruction conversion on an instruction set adopted by the artificial intelligence model; the artificial intelligence model issues an instruction to the weighted value central processing unit through the converted instruction set; and after receiving the instruction, the weighted value central processing unit averagely distributes a plurality of execution threads to execute a plurality of programs indicated by the artificial intelligence model according to the weighted value.

A search query to provide a search result may be received, which identifies source and target nodes and an application for generating the search result. The application accesses a data structure holding landmark nodes, which store a shortest path tree in the form of a set of parent links. Each parent link can identify an adjacent vertex node in a shortest path between each node in the data structure and the landmark node. The location of the source node and the target node in the shortest path trees may be identified to the landmark node. For each landmark node, using the identified locations of the target node and source node, a measure of distance between the source node and the target may be generated. The landmark node with the shortest distance may be determined. A search result related to the shortest path tree of that landmark node may be provided.

The invention discloses a SIG mesh-based broadcast bearer layer message filtering strategy method, the method is based on the bluetooth SIG mesh mesh network, and the mesh mesh network includes a bearer layer, a network layer, a transport layer and an access layer, The bearer layer transmits data between mesh nodes and passes it to the network layer, and the mesh node receives broadcast messages at the bearer layer, and compares them with preset queues to realize the mesh Filtering of broadcast packets in a mesh network reduces RAM consumption and CPU execution time. The invention optimizes the consumption of RAM, CPU execution time, and BLE scanning time due to the interference of non-mesh broadcast packets in the air, and optimizes the consumption of RAM and CPU execution time caused by repeated data packets.

The present invention provides a method and system for task analysis and resource allocation based on a formal model, including: step S1: a task analysis module analyzes task requirements, and formally expresses the task requirements as a task workflow; step S2: a resource allocation module Combine the formalized task workflows to obtain a combined task workflow; step S3: abstract the resource capabilities of the mobile terminal device into a meta model, and perform service registration in the cloud; step S4: based on the combined task workflow and For the service registered in the cloud, a heuristic algorithm is used to perform task scheduling and matching; step S5 : the mobile terminal device executes the task according to the scheduling matching result.

The invention relates to a cloud computing load balancing task scheduling method based on cellular automata. The initial solution of task scheduling is obtained based on the heuristic priority allocation strategy of the earliest completion time; then the initial solution of task scheduling is optimized based on cellular automata and greedy principle. solution, thereby minimizing the latest completion time of the overall task and improving the load balancing performance of the virtual machines. Adopting the cloud computing load balancing task scheduling method based on cellular automata of the present invention not only effectively improves the efficiency of task scheduling, realizes the load balancing of the cloud computing platform, but also significantly reduces resource idle time and improves resource utilization benefit, and has better practicability.