11 results about "Graph coloring" patented technology

The present application relates to a kind of based on position dynamics adaptive anisotropic cloth simulation method, system, electronic equipment and storage medium, belong to computer graphics physics simulation field.The method includes setting cloth grid;Topology relationship of grid is calculated, and the set of point edge surface is formed;According to Newton's law, the linear system of external force is constructed, and change coordinate is solved using conjugate gradient method;Using graph coloring method, data dependence is decoupled to grid, using the method of position dynamics, strain constraint iteration parallel computing is carried out to change coordinate, and the coordinate correction amount after cloth motion deformation is obtained;According to the revised coordinate, grid tensor field is calculated, grid dynamic refinement and coarsening are carried out, and non-uniform anisotropic cloth grid is formed.The present application uses position dynamics method to calculate dynamic non-uniform grid, simulates the anisotropy of cloth deformation, retains the dynamic behavior of cloth wrinkle under fine scale, is quickly solved on the basis of not losing simulation details, satisfies real-time.

A distributed inference communication method of a super-large-scale graph neural network comprises the following steps: dividing a global topology to be processed into a plurality of local sub-domains matched with the number of GPUs, and determining a local master node and a boundary master node allocated to each GPU; calculating coloring priorities of nodes, and performing parallel graph coloring based on the coloring priorities to construct a global independent set sequence; mapping the global independent set sequence to the local sub-domains of the GPUs to generate a local task slice sequence, which comprises boundary node slices and internal node slices, evaluating communication load and calculation load of the slices, and taking the internal node slices as calculation fillers to be executed in the communication waiting period of the boundary node slices; and each GPU executes an inference task by using a multi-stream asynchronous pipeline constructed locally according to the local task slice sequence. The method can improve the throughput of distributed inference.

The invention provides a high-energy-efficiency digital-analog hybrid CMOS Boolean satisfiability problem solver chip which comprises a solver core, an SRAM peripheral read-write drive circuit and a parallel-to-serial read-out circuit, the SRAM peripheral read-write drive circuit is connected with a processor core and used for reading and writing data, the parallel-to-serial read-out circuit is used for reading a final solution, and the solver core comprises a 218 * 50 processing unit array. The processing units in each column are interconnected through charge sharing lines so as to complete accumulation operation in a charge domain. The processing units in each row are interconnected through TCAM matching lines, and the word and sentence evaluation operation is completed. In addition, a graph coloring circuit is also integrated for grouping Boolean variables. In each iteration, the Boolean variables in the same group can be updated at the same time, so that the parallel computing capability of the solver is improved. The energy efficiency is improved by using a digital-analog hybrid calculation mode, and the degree of parallelism is improved by using a packet updating strategy.

The invention discloses a traffic scheduling method and system in a distributed training task scene, which are used for distributed training tasks between a first device end and a second device end. The method comprises the following steps: according to a communication demand of a distributed training task, performing bipartite graph abstraction and graph coloring processing on a first device end, a second device end and network resources between the first device end and the second device end in sequence, establishing a communication model, and establishing a traffic scheduling optimization problem in combination with a first traffic scheduling target and a second traffic scheduling target; the first flow scheduling target selects paths flowing through different core switches for communication flows with the same source address or target address in any communication stage of a single task, and the second flow scheduling target selects paths flowing through different core switches for communication flows with the same source address or target address among different tasks; and solving a traffic scheduling optimization problem to obtain a path set and complete routing. According to the method, the flow conflicts in the tasks and among the tasks in the distributed training task scene are reduced.

The invention relates to a dynamic routing method based on storage and forwarding in a space division multiplexing elastic optical network, and belongs to the technical field of communication. The method comprises the following steps: firstly, acquiring the current state of a time-shifting multilayer graph, performing routing in the time-shifting multilayer graph, then determining the total occupied frequency slot number according to a path distance and a modulation format, and selecting a transmission path of a request in combination with the balance of multi-dimensional resources; then grouping and grading the fiber cores by using a graph dyeing method, and selecting the fiber cores for each link by considering crosstalk of the fiber cores and frequency slot use conditions according to a fiber core selection method designed in the invention; secondly, designing a frequency spectrum selection method, and selecting a proper frequency spectrum block for the request by comprehensively considering fragments of time domain and frequency domain dimensions; and finally, carrying out allocation in the time-shifting multilayer graph, and updating topology and resource use states until all requests are allocated. The storage and forwarding process can be simplified while the network capacity is improved, and the blocking rate is improved.

The application provides an optical wave signal distribution method, device and system, which are applied to the technical field of optical communication. The method converts the distribution problem of optical wave signals into a coloring problem of a conflict graph corresponding to the optical wave signals, and then uses a related algorithm model to solve the graph coloring problem to obtain a coloring scheme of the conflict graph. The coloring scheme of the conflict graph has a corresponding relationship with an optical wave signal distribution scheme of the optical wave signals, and accordingly the optical wave signal distribution scheme can be obtained. Based on the method, the distribution of the optical wave signals in an optical switching architecture can be automatically calculated by a device, without human planning, and the planning efficiency of the optical wave signal distribution scheme in the optical switching architecture can be greatly improved.

The invention relates to the technical field, in particular to a general nameplate format setting and printing method, which comprises the following steps of: comparing a code scanning entry item, a server database and a local data source, removing repeated contents by utilizing uniqueness retrieval, and executing sorting and restructuring between an attribute field and a content field, so as to obtain a general nameplate format; a data sequence structure is more compact and has logic coherence, coordinate values, font sizes and rotation angles are uniformly written into a real number string, and different combinations are iteratively screened under three quantitative indexes of an overlapping rate, a blank rate and an alignment offset by combining exchange and disturbance operations of a genetic algorithm, so that the accuracy of the real number string is improved. The space utilization rate is maximized, information shielding is avoided, boundary parameters are converted into nodes, a graph coloring algorithm is adopted to allocate colors for intersection areas, meanwhile, the positions of titles and parameters are adjusted according to a priority table, adjustment is stopped when the number of conflict edges is lower than a threshold value, and it is ensured that the overall structure meets readability and logic orderliness.

To provide a vibration neural network circuit device for solving a graphcoloringproblem by allowing a plurality of oscillators coupled by a capacitor or a resistor to detect an edge or to indicate colors of at least two graphs for an input graph.SOLUTION: In the vibration neural network circuit device 700 using a gated diode-based oscillator, phase differences between the voltage VIN1 and the voltage VIN2 may occur due to time differences between the voltage VOUT1 and the voltage VOUT2 input to at least two gated diodes constituting at least two oscillators 701 and 702, and thus colors of at least two graphs with respect to input graphs of a graph coloring problem may be represented based on the phase differences over time.SELECTED DRAWING: Figure 7A

The application belongs to the technical field of electric digital data processing, and relates to a complex construction node steel bar arrangement logic optimization method. Steel bar geometric parameters and anchoring constraint relationships are extracted, a directed acyclic graph is constructed with steel bar sections as nodes, spatial occupation intersections as edges, and construction sequences and construction specifications as weights; spatial arrangement conflicts are converted into loop identification problems in the graph, and graph coloring processing is used to divide conflict edges into levels and give avoidance priorities; based on topological sorting traversal, bending avoidance directions and offset rules are determined according to coloring values; offset compensation calculation is performed through topological reverse sequence, offset values are reversely transmitted to non-conflict steel bar sections along the edges, and arrangement coordinates and bending node logic sequences are output. The application converts serial collisions into graph structure operations, overcomes chain collisions and local deadlocks caused by serial avoidance, and eliminates space occupation caused by avoidance.

The invention discloses a clothing care machine control method and system, and belongs to the technical field of clothing care. The method comprises the following steps: collecting appearance images of put clothing, and extracting color texture distribution, glossiness, edge softness and other material characteristic factors; in combination with a preset clothing material classification database, identifying the material category of the clothing and the corresponding temperature, humidity and wind speed sensitive factors; generating a nursing parameter template based on the identification result; conflict detection is carried out on the nursing parameters of the multiple clothes, and a multi-clothes nursing coordination map is constructed; dynamically segmenting the nursing time axis through a graph coloring algorithm to generate a conflict-free nursing sequence; environmental parameters in the box body are monitored in real time in the nursing process, and the target nursing state is dynamically adjusted; according to the invention, intelligent identification and fine nursing of clothes made of various materials can be realized, clothes damage caused by improper parameter setting is effectively avoided, and the nursing effect and the equipment energy efficiency are improved.