44 results about "Graph partition" patented technology

The invention discloses a method which utilizes a graph partition method to automatically extract a multi-document summarization, and the method comprises the following steps that: the sentence boundary dividing is carried out, and the document is expressed by the divided sentences; the sentences are expressed into vectors, the similarities among each two sentences are calculated to compose a sentence incidence matrix, which is reduced according to the appointed threshold value, at the same time, the normalized treatment is carried out; the crawling of the implied logical topic of a topic is introduced into the multi-document summarization, and a document set is divided into different implied sub-topics according to the topic, thereby the summarization task is changed into the selection and the extraction processes to the sub-topics; by applying the graph partition method, the importance degree of the sub-topic of the sentences is ensured from the global characteristics, and the low redundancy of the contents among the different sub-tops is ensured from the local characteristics, thereby effectively improving the quality of the summarization.

Methods, machines, and stored instructions are provided for partitioning a graph of nodes into clusters of nodes by iteratively excluding edges in the graph. For each node of at least a subset of nodes in the graph, a graph partitioning module determines whether to exclude edges for the node and, if so, selects for exclusion edge(s) to at least a subset of the node's neighbor(s). The module selects edge(s) to the node's neighbor(s) for exclusion based at least in part on a degree of overlap between the node's neighbor(s) and neighbor(s) of the node's neighbor(s). For any subset(s) that are yet not sufficiently partitioned into clusters, the module repeats the step of determining whether to exclude edges and, if so, selecting nodes for exclusion, and determining whether or not the nodes are sufficiently partitioned. Subset(s) of nodes that are already sufficiently partitioned may be skipped during the repeated steps.

The invention is suitable for the technical field of image processing, and provides a method and a device for compressing a composite image. The method comprises the following steps: S1, partitioning the composite image according to image features to obtain a plurality of partitioned areas, wherein the image features include color and grayscale, and the partitioned areas include a text / graph area and an image area; S2, compressing the partitioned areas by adopting a corresponding encoding way respectively. In the method, the composite image is partitioned into text / graph partitioned areas and natural image partitioned areas according to information of different features included in the composite image, so that corresponding compression methods can be adopted for different partitioned areas, and high compression ratio is ensured. Meanwhile, the text / graph areas are clear and readable after being reconstructed.

A method, computer program product, and a data processing system for determining an edge for inversion in a cyclic compound directed graph. A plurality of graph nodes of a graph are evaluated for generation of a first node order subset pair. A determination that the graph requires a recursive evaluation for completing node ordering of the plurality of graph nodes is made. The graph is divided into a plurality of graph partitions, and respective node order subset pairs are generated for the graph partitions.

A system for executing a graph partitioned across a plurality of reconfigurable computing units includes a processing node that has a first computing unit reconfigurable at a first level of configuration granularity and a second computing unit reconfigurable at a second, finer, level of configuration granularity. The first computing unit is configured by a host system to execute a first dataflow segment of the graph using one or more dataflow pipelines to generate a first intermediate result and to provide the first intermediate result to the second computing unit without passing through the host system. The second computing unit is configured by the host system to execute a second dataflow segment of the graph, dependent upon the first intermediate result, to generate a second intermediate result and to send the second intermediate result to a third computing unit, without passing through the host system, to continue execution of the graph.

The invention provides a tracking area list management method based on overlapping community detection in a small cellular network. For user location management in the small cellular network, the signaling expenditure of a network system is optimized, and a determination method of a TAL is provided. For the giving of the small cellular network mode after TA planning, the general idea of the methodprovided by the invention is as follows: firstly counting tracking area location updating and paging data of the user at the specific time frame and in the specific region; and then modelling a TAL management method as a graph partition problem, and giving a linear planning model; and finally giving a TAL structure by applying an overlapping community detection algorithm based on the game theory.

The invention discloses a multi-task external memory mode diagram processing method based on I / O scheduling, includes streaming partitioning graph data to obtain graph partition, evenly placing graphpartition in multiple external storage devices, selecting target external storage devices from multiple external storage devices based on I / O scheduling, and taking graph partition in the target external storage device that has not been accessed by graph processing task as designated partition; Judging whether the synchronization field of the designated partition is not mapped into the memory according to the synchronization field of the designated partition, if so, mapping the designated partition from the external storage device into the memory, and updating the synchronization field of thedesignated partition; Otherwise, the graph partition data is accessed directly through the address information mapped to memory by the specified partition. Through I / O scheduling, the invention selects the external storage device with the least number of tasks to access, thereby controlling the sequence of accessing the data of the external storage diagram partition and balancing the I / O pressure.By setting the synchronization field to realize the data sharing of graph partition, the repeated loading of the same graph partition is reduced, so as to reduce the total I / O bandwidth and improve I / O efficiency.

The invention relates to a pedestrian trajectory prediction method based on a GP-CNN, and aims to solve the core problem of trajectory prediction in automatic driving, namely how to design a model to better capture associated interaction information to improve prediction precision and safety of an automatic driving automobile. According to the invention, a pedestrian trajectory prediction model applied to a complex scene is designed, based on the GP-CNN, interaction features of the scene are extracted by using a mode of combining two channels to serve as input, and information of time-domain features of the pedestrian trajectory is specially processed, meanwhile, forward and backward propagation of the prediction trajectory is smoother through residual connection, and then a pedestrian interaction prediction trajectory is generated through a trajectory prediction network.

The invention provides a Method for estimating optimal power flows in power grids using consensus-based distributed processing. The method estimates an optimal power flows (OPF) in a power grid, which is represented as a graph partitioned into virtual sub-graphs, each including at least one bus, and associated with agents that measure local variables and updates consensus variables (CV). The consensus variables of adjacent virtual sub-graphs are exchanged and updated using the agents. An OPF problem is solved for the virtual sub-graphs using the agents based on the CV and the local variables. The exchanging and the solving are iterated until a termination condition is satisfied, when the optimal OPF is outputted for each virtual sub-graph.

The invention discloses a multi-scale fusion network simulation task mapping method in a heterogeneous environment, which solves the deployment problem of a virtual network topology in a heterogeneous computing cluster environment. The steps of the method include: reading into a heterogeneous computing environment; Mark the edge routers and host nodes in the topology as lightweight virtualization mapping areas, and mark the remaining nodes as converged virtualization mapping areas; according to the throughput threshold of each server, use lightweight virtualization mapping for lightweight virtualization mapping Nodes in the area; calculate the load balance parameters of the remaining servers, use the multi-level graph partition algorithm to allocate nodes in the fusion mapping area, judge whether the server is redundant, and use different optimization algorithms to optimize and map reasonably according to the results. The invention ensures load balance between computing clusters, reduces remote communication, improves the performance of large-scale network simulation, and has good scalability and scalability for large-scale network topology, and can be used for various network research and experimental networks.

The invention discloses a storage optimization-based distributed graph processing method, and belongs to the field of graph calculation. The method comprises the following steps of: carrying out data partitioning in a data pre-processing stage; distributing graph partition data; starting data iteration processing; updating message transfer; making a work node extension decision; and ending the data processing. According to the method, a consistency hash algorithm is proposed to partition and store graph data, and an external storage mode-based distributed graph processing system is designed and realized; and by utilizing a dynamic storage optimization strategy, graph data processing load balance is realized according to partition storage of load adjustment graphs, so that the graph data processing speed is improved, the problem that the overall performance is reduced as hotspots are caused by load imbalance in the graph data processing process in the prior art is solved, and then the graph processing performance is improved.

The present invention relates to a game-based method and system for streaming-graph partitioning, the method comprises: partitioning a streaming graph using one or more processors, the one or more processors being configured to: read an edge streaming having a predetermined number of edges in an unpartitioned area of the streaming graph as a sub-graph; based on a first pre-partitioning model, pre-partition the edges of the sub-graph to at least two partition blocks as an initial state of a game process; and sequentially select an optimal partition block for each edge of the sub-graph through the game process until the game process becomes convergent, the disclosed method and system can partition streaming graph using local information only, without loading the whole streaming graph into the memory, thus have good scalability and support dynamic graph partitioning; the disclosed partitioning method and system can provide better partitioning results.

The present application relates to a method, system, device, and storage medium for counting the number of vertices and edges in a graph database, wherein the method includes: obtaining a task request for counting the number of vertices and edges in the graph space, and checking the preset preparation queue in the preparation queue according to the task request. Task, returns the ID of the preset preparatory task, writes the information record of the preset preparatory task in the system table of the preset preparatory task according to the ID of the preset preparatory task, and writes the graph in the system table of statistical information according to the ID of the graph space For spatial information records, the preset preparatory tasks are grouped according to the Leader node to which the graph partition belongs, and several preset execution tasks are obtained, and the preset execution tasks are run through the storage node. It solves the problem of how to reduce the waste of computing resources and improve the efficiency of real-time statistical data, and realizes an efficient parallel strategy for point-edge statistical tasks based on the job mechanism and task mechanism in Nebula Graph.

The invention discloses a CAD data processing device and a CAD data processing method. Specifically, the invention relates to the field of computers, the device comprises a graph scanning module, a graph partitioning module, a graph extraction module, a graph locking module, a graph merging module and a graph unlocking module, the graphic scanning module comprises a graphic scanning unit; the graph partitioning module comprises a graph partitioning unit; wherein the graph extraction module comprises a graph layer new establishment unit and a graph copy unit, the graph locking module comprises a graph locking unit, the graph merging module comprises a graph merging unit and a graph layer deletion unit, and the graph unlocking module comprises a graph unlocking unit and a graph storage unit. A graph is scanned through the graph scanning module, partition is conducted through the graph partition module, a graph area needing to be modified is extracted through the graph extraction module, an area not needing to be modified is locked through the graph locking module, and the safety of the system is higher on the whole.

Embodiments of the invention relate to data archiving on storage medium such as magnetic tapes based on graph partitioning. One aspect of the invention concerns a method that comprises representing a file system as a graph where each node of the graph represents a file and each edge represents file chunks common to the files corresponding to the nodes connected by the edge. The graph is repeatedly partitioned into disjoint sub-graphs. If the files and duplicate file chunks associated with a sub-graph fit into a data storage medium, then the files and duplicate file chunks are stored in the medium. Otherwise, the method may partition the sub-graph into smaller disjoint sub-graphs taking into consideration of degrees of connectivity of the nodes.

The invention discloses a block-based subgraph construction and distributed graph processing method. The block-based subgraph construction method comprises partitioning a graph, relabeling summits in the graph, associating intervals with shards, partitioning blocks and constructing subgraphs, and employs a heuristic and lightweight SGP (Streaming Graph Partition) method to partition a graph, and a user-defined heuristic function to designate the summits to P subgraphs in order; the method possesses the characteristics of high performance and low edge cutting rate, and considers subgraph work load balance. According to a distributed graph processing system, after partitioning, subgraphs perform communication by employing a whole memory block as a unit; compared with the fine grit communication mode of the prior art, the distributed graph processing system exchanges data based on a memory block, and can fully utilize network bandwidth and reduce communication time.

The invention discloses a large-scale entity alignment method based on bidirectional graph partition and reciprocal reasoning. The method comprises the following steps: acquiring data of a source knowledge graph and a target knowledge graph; performing graph partition on the source knowledge graph and the target knowledge graph to obtain a group of sub-graph pairs of source sub-graphs and target sub-graphs; obtaining joint entity representation in all the sub-graph pairs by utilizing a representation learning model of an entity structure; performing reciprocal alignment reasoning on sub-graphs in the sub-graph pairs to obtain alignment results of the sub-graph pairs; and aggregating the alignment results of all the sub-graph pairs to obtain an alignment result of the source knowledge graph and the target knowledge graph. According to the method provided by the invention, through bidirectional partition and aggregation, the final sub-graph pairs can have a more complete structure and more seed entity pairs, so that a more accurate entity alignment result can be generated; and compared with traditional direct alignment reasoning, two-stage reciprocal modeling of reciprocal reasoning can obtain a better alignment effect.