36 results about "Network decomposition" patented technology

The invention relates to the technical field of chip testing, and discloses an LED driving chip testing method, device and equipment and a storage medium, and the method comprises the steps: carrying out the multi-point temperature excitation and constant-current output synchronous collection of an LED driving chip, and obtaining electrothermal coupling test data; performing thermal resistance network decomposition on the temperature field in the chip to obtain temperature distribution characteristic parameters; performing nonlinear function fitting on the temperature-constant current coupling relation to obtain an electrothermal coupling transfer coefficient; performing four-dimensional coupling calculation on the temperature distribution characteristic parameter and the electrothermal coupling transfer coefficient to obtain four-dimensional coupling response characteristic data; and partial differential sensitivity calculation is carried out based on the four-dimensional coupling response characteristic data, and an electrothermal coupling sensitivity analysis result is obtained.According to the method, the problem of test errors caused by time desynchrony in a traditional separated temperature-current test is solved, quantitative sensitivity analysis of electrothermal coupling parameters is achieved, and the sensitivity of the electrothermal coupling parameters is improved. Therefore, the test accuracy of the LED driving chip is improved.

The invention relates to the technical field of industrial equipment intelligent simulation, and discloses a simulation model construction method and system, and the method comprises the following steps: 1, analyzing the electrical control logic of industrial equipment, and converting the electrical control logic into a mixed integer nonlinear programming problem containing a discrete Boolean variable and a continuous variable; 2, modeling the physical field based on a tensor network decomposition algorithm, and generating a compressed low-rank tensor network model; 3, dynamically adjusting parameters of the tensor network model by adopting a hierarchical optimization strategy, wherein the hierarchical optimization strategy switches a global search algorithm and a local optimization algorithm according to an error convergence state; 4, allocating and executing symbol calculation, tensor network operation and parameter optimization tasks through a heterogeneous calculation system; and 5, collecting production line sensor data in real time. According to the method, different physical field sub-models are dynamically activated through Boolean variables, and the accuracy of joint simulation in a complex industrial scene is remarkably improved.

The invention discloses a roadbed pavement test and data acquisition system under the action of moving load, which comprises the following steps of: acquiring high-precision pavement unevenness data, decomposing features through a space-time convolution variational self-encoding network and inhibiting noise; based on the normal stiffness and the morphological gradient, the density of the sensor is adjusted by adopting dynamic grid division, and multi-modal data is fused by utilizing Bayesian weight. And adaptively extracting signal details by applying multi-scale transformation, and designing a filter to enhance the damage signal and optimize the structure entropy. And constructing a multi-field coupling grid control equation to realize dynamic topological optimization reconstruction. And establishing a damage evolution state space, and training the prediction model through a multi-stage reinforcement learning framework in combination with experience playback. And constructing a digital twin mapping model, integrating Bayesian optimization and reinforcement learning, and establishing a virtual-real feedback channel to realize dynamic decision optimization. According to the invention, innovative improvement on a traditional test method and a data acquisition system is provided, and meanwhile, complex nonlinear dynamic behaviors can be accurately simulated and analyzed.

The invention discloses a personalized cross-domain recommendation method and system based on federal learning, and the method provides a personalized cross-domain recommendation service for a user on the premise that original data of interaction between the user and an article and user parameters are kept locally. The method comprises two stages of federated training: stage 1, intra-domain users cooperatively train a single-domain score prediction model by using a neural collaborative filtering method; in the second stage, overlapping users of the two domains cooperatively train a migration module based on a multi-layer neural network to capture a mapping relation represented by potential user features between the two domains; besides, each layer of network of the cross-domain recommendation model is decomposed into a base vector and a personalized vector which respectively represent common knowledge among different users and unique knowledge of the users, and a local model obtained by final training can provide personalized recommendation services for registered users in a target domain, so that the recommendation efficiency is improved, and the user experience is improved. Meanwhile, the global model obtained through training provides effective initial recommendation for new users in the target domain, and the cold start problem in a recommendation algorithm is effectively relieved.

The invention belongs to the technical field of image processing, and particularly relates to a non-uniform low-illumination image enhancement method and system. The method comprises the following steps: step 1, constructing a network model: constructing a non-uniform low-illumination image enhancement network model, wherein the model comprises a Retinex network, an illumination image enhancement network, a reflection image enhancement network and a multi-scale enhancement network; step 2, preparing a data set: dividing a training set and a test set by using an LOL-v1 data set, and preprocessing the training set and the test set; and performing fine adjustment on the model by the LOL-v2 data set. According to the invention, based on deep learning, an efficient two-way feature enhancement network is designed and is used for improving the performance of image features. On the basis of Retinex network decomposition, an input image is decomposed into an illumination image component and a reflection image component, and the characteristics and channel dimensions of the illumination image component and the reflection image component are modulated respectively, so that the characteristics of a low-illumination image are effectively enhanced. According to the method, the quality and definition of the image can be remarkably improved, and illumination and reflection details are brighter.

The invention discloses a wireless networking fault positioning method, device and system based on a binary tree algorithm, and relates to the technical field of communication network security, and the key points of the technical scheme are as follows: obtaining alarm information of a monitored network; judging whether the monitored network has a single-link fault or a multi-link fault according to the alarm information; if the fault is a single link fault, positioning the position of a fault point based on the topological structure of the monitored network, the alarm information and a pre-acquired binary decision tree model; if a multi-link fault occurs, decomposing the monitored network into a plurality of binary tree sub-networks; and for each binary tree sub-network, positioning the position of the fault point based on the topological structure of the binary tree sub-network, the alarm information and a pre-acquired binary decision tree model. By adopting the scheme provided by the invention, the timeliness and the accuracy of networking fault positioning are improved, and the safety and the reliability of networking communication are ensured.

The invention discloses a heterogeneous network embedding method based on dynamic heterogeneous network decomposition and a hierarchical space-time attention mechanism. The method comprises the following steps: dynamic network decomposition: decomposing a heterogeneous network sequence into independent subgraph sets according to edge types; multi-order node attention aggregation: for each sub-graph, fusing information of nodes and different edge types; semantic level attention is embedded, wherein global semantic embedding is generated in combination with global relation type weight and meta-path long-short-term memory network coding; time dynamic modeling: respectively calculating a time attention weight and a Monte Carlo sampling approximate Horkes intensity, and then aggregating and splicing results of the time attention weight and the Monte Carlo sampling approximate Horkes intensity to obtain a final embedding result. According to the method, through cross-granularity semantic modeling and efficient time sequence dependence learning, an innovative solution is provided for node representation of the dynamic heterogeneous network.

The invention discloses a moving target detection method based on low-rank full-connection tensor network decomposition and a total variation model, and belongs to the technical field of computer vision and real-time video analysis. Aiming at the problems of poor real-time performance, dynamic background misjudgment, serious noise interference and the like caused by neglect of space-time relevance, fixed rank constraint and high calculation complexity of a traditional method in a complex dynamic scene, the following technical scheme is provided: through full-connection tensor network (FCTN) decomposition, a common rank is utilized to realize accurate modeling of a dynamic background; interference from background leakage to foreground is reduced; incremental updating of time core parameters is carried out in combination with a sliding window strategy, and a historical space core is fixed to reduce calculation complexity; total variation (TV) regularization is combined to optimize a low-rank-sparse separation process, the continuity of a space gradient and a time gradient is restrained, the contour integrity of a moving target is enhanced, and noise is suppressed.

The invention discloses a time sequence and congestion collaborative optimization-oriented GPU (Graphics Processing Unit) parallel acceleration global wiring method, which comprises the following steps of: according to a given netlist, dividing a super-large network by adopting a Kruskal algorithm in combination with a lookup set, and constructing a time sequence propagation path of the super-large network; performing network decomposition based on the timing margin estimation of the pins; calculating the time sequence weight of the two-pin network by adopting a time sequence weight calculation method based on an Elmore delay model; performing path cost calculation by adopting a cost function which comprehensively considers the time sequence and the congestion cost; executing two-level GPU parallel kernel acceleration mode wiring; optimizing the network topology by adopting a delay-aware pin connection improvement technology; the invention relates to a congestion-driven GPU (Graphics Processing Unit) accelerated routing and routing strategy for executing non-critical networks. According to the method, a high-quality wiring result with balanced time sequence congestion can be quickly obtained, the time sequence performance is effectively improved, and the requirement of a current super-large-scale high-performance circuit design wiring stage can be met.

The invention relates to the field of image enhancement, in particular to a low-light image enhancement method, which comprises the following steps of: constructing a training data set by using an underexposure low-light image and an aligned low-light image in a plurality of scenes; constructing and training a low-light image denoising network in combination with a preset low-light loss function; the low-illumination loss function comprises loss for measuring the difference between two pairs of transformed low-light images, loss for measuring the difference between the denoised and down-sampled image and the target image, and structural similarity loss; constructing and training a decomposition network based on the de-noised image output by the LD-Net; the loss function of the decomposition network comprises reconstruction loss, reflectivity consistency loss, illumination smoothness loss and decomposition loss; and a de-noising Retinex decomposition model is generated according to the LD-Net and the decomposition network, and is used for carrying out image enhancement operation on the to-be-processed image based on the illumination correction factor, so that the model has less parameter quantity through a simple network design, and meanwhile, the model has better detail recovery effect and enhancement effect.

The invention discloses a flexible job shop scheduling method based on a generative adversarial training framework. The method comprises the following steps: establishing a Markov decision process model for flexible job shop scheduling, and completing the design of a state space, an action space and a reward function; collecting expert scheduling tracks through a plurality of algorithms, and carrying out data cleaning and standardization processing; constructing a state encoder based on a graph attention network, mapping a scheduling environment state into low-dimensional vector representation, and designing a hierarchical strategy network to decompose a scheduling decision task; constructing a value network to provide stable value estimation so as to accelerate a reinforcement learning process, and constructing a discriminator network to guide a strategy search direction by generating an imitation reward; and finally, hybrid training based on near-end strategy optimization and generative adversarial imitation learning is executed, and through adversarial training and strategy gradient optimization, an intelligent agent learns to obtain a high-performance scheduling strategy with expert empirical performance and environment adaptability.

The application discloses a kind of based on the transport hub maintenance scheduling efficiency optimization method of deconstruction of variable topology super network, comprising the following steps: step one, the maintenance element involved in the current real-time maintenance scheduling scene of transport hub and its interrelated relationship mapping modeling is modeled into supergraph network model;Step two, the basic uncertainty characteristics of supergraph network CNN topological structure are characterized;Step three, according to the supergraph network model, the evaluation value of measurement index is calculated;Step four, the three indexes calculated are compared with the reasonable value interval range of historical scheduling result, and the rationality of current scheduling state is judged;Step five, based on the measurement index, the maintenance scheduling efficiency of transport hub is improved.The beneficial effects of the present application are that the execution efficiency, resource collaboration degree and dynamic robustness of maintenance scheduling can be effectively improved through super network modeling and quantitative analysis, and bottleneck positioning and optimization are realized.

The application discloses a social network link prediction method and device, the method comprises the following steps: decomposing a social heterogeneous network into a plurality of account association sub-views through a social network meta-path, using a graph convolution network to represent the accounts in each account association sub-view, fusing the features of the multiple views through an attention mechanism to generate an account feature vector, using the account feature vector to construct a social network link prediction model, and calculating a link score to realize link prediction. By using the social network meta-path to extract the multi-dimensional association relationship between the accounts, more heterogeneous information between the accounts is considered, and the attention mechanism fuses the features under the multiple account association sub-views to automatically calculate the contribution of various relationships to the link prediction effect, thereby solving the problem that the utilization rate of the social network heterogeneous association relationship is low in the prior art, and the link prediction accuracy is not high.

The invention provides a parallel structured network subdivision system suitable for an FDTD algorithm, and belongs to the technical field of electromagnetism. Surface fitting is conducted on each triangular surface element through a surface fitting module, a triangular surface element complexity list is generated, the load process of each triangular surface element in the complexity list is dynamically allocated based on a greedy algorithm, and a global interface file containing surface data is generated; carrying out internal grid filling on the target object subjected to surface fitting through a medium filling module, creating a Cartesian structure for each grid plane process needing to be filled, and carrying out filling in a parallel mode to generate a grid model; and realizing cross-module data transmission through the global interface file, and outputting a grid file containing a grid model. According to the method, the problem of resource idleness caused by the difference of the calculated amount of the triangular surface elements is effectively solved, the parallel efficiency is remarkably improved, the surface fitting module can achieve the acceleration effect close to linearity, and the method is particularly suitable for large-scale geometric models with complex structures.

This invention relates to the field of off-grid power supply systems with multi-energy coupling, and discloses a hardware-in-the-loop (HIL) simulation method for an off-grid power supply system with DC coupling of wind, solar, and hydrogen storage. The method includes the following steps: building a real-time simulation model of the off-grid power supply system; designing and establishing communication links; developing hardware and software for equipment controllers; and building and testing a hardware-in-the-loop real-time simulation platform. This invention's HIL simulation method for off-grid power supply systems with DC coupling of wind, solar, and hydrogen storage improves modeling through model simplification, network decomposition, and electrical decoupling, achieving high-precision real-time simulation of the off-grid power supply system.

The invention discloses a distributed multi-agent self-adaptive optimization incremental power distribution network power voltage balancing method, which comprises the following specific steps: carrying out data monitoring by a voltage and current sensor, acquiring real-time voltage and current data of each feeder line of a system, calculating real-time power on each sub-network node by agents through the data, and calculating the real-time power of each sub-network node according to the real-time power of each sub-network node; calculating reference power by using real-time power data of the current converter and an adjacent agent, and realizing decoupling control of voltage by the current converter through the data; a radiation type system is considered, a virtual voltage source-sub-network decomposition model is constructed, the system is decomposed into a plurality of sub-networks, a network construction type power supply is decomposed into two virtual voltage sources, and each sub-network comprises the virtual voltage sources on the two sides, an internal load and a network following type power supply; and constructing a cost optimization function of each sub-network, carrying out iteration based on a spherical search analysis method to continuously obtain optimal voltage and power points, and carrying out multiple iterations to realize the optimal solution of the voltage and power of the system.

In one implementation, we model information content in a signal as a synthesis of global and local information. The global information is common and shared for all natural signals, and this can be learned from a big data set. However, the local information is specific for each signal. By means of the synthesis property of a neural network, the INR network is decomposed into a head layer and a tail layer, the head layer is responsible for global information, and the tail layer is responsible for local information. The weights of the header layer are approximated based on a dictionary, and the dictionary is learned from a big data set and known to both encoders and decoders. Therefore, we need to send only the weight of the tail layer plus some additional information about the head layer.

The invention discloses a line loss diagnosis method based on multi-source data, diagnosis equipment, a medium and a program product, and relates to the technical field of intelligent power distribution network digital operation and maintenance. According to the invention, accurate diagnosis and abnormal positioning of the line loss of the transformer area are realized. According to the topology calculation unit division, a complex transformer area network is decomposed into manageable basic units, so that the line loss analysis is more refined. The unit theoretical line loss-load characteristic curve model established by regression fitting provides a scientific reference for anomaly detection. The theoretical line loss rate is obtained by mapping a future period load curve obtained by the prediction model, and the theoretical line loss rate is compared with the real-time line loss rate to form an anomaly detection process from macroscopic to microscopic. And accurate positioning of the abnormal unit is completed through deviation feature recognition. Fine management from the whole transformer area to a specific unit is realized, line loss management is improved from post-event statistics to an active mode of pre-event prediction and in-event intervention, and the detection efficiency and accuracy of line loss abnormity are improved.

The invention discloses a non-intrusive load decomposition method based on a gating circulation unit, relates to the technical field of power system planning and analysis, and is used for solving the problem of insufficient decomposition precision of a difference method. Comprising the steps of load decomposition, generation of a GAN, generation of a GRU and construction of a GRU-GAN network decomposition model. According to the method, a GAN (Generative Adversarial Network) framework is utilized to reconstruct a load decomposition problem, and the load decomposition problem is defined as a denoising problem. The GAN cannot extract the time dimension characteristics of the current sequence, and the GRU has the advantage of being suitable for processing long sequence data, and can effectively store and utilize historical information and time sequence characteristics in the time sequence data. Therefore, the GAN and the GRU are combined, that is, the input of the generator G of the GAN is used as the output of the GRU, so that the improvement of the original GAN network is realized, and the GRU-GAN load decomposition model is obtained.

This invention discloses a method for decomposing complex networks based on graph contrastive learning and multi-hop aggregation, comprising the following steps: collecting the number of neighbor nodes, connecting edges, and average clustering coefficients of a complex traffic network to construct a network decomposition model; inputting the original graph into a role graph generation module to obtain a role graph; inputting the original graph and the role graph into a multi-view representation learning module to obtain multi-angle graph representations of the role graph and the original graph; obtaining an importance score for each node; calculating and optimizing the joint loss function to train the network decomposition model; decomposing the traffic network to obtain the importance values ​​of traffic nodes in the traffic network, and setting stronger security measures for traffic nodes with high importance and weaker security measures for nodes with low importance. This invention uses intra-graph contrastive learning and cross-contrast learning to improve graph representation performance and proposes a multi-hop aggregation mechanism to predict node importance by combining multi-hop neighbor information, achieving high-performance network decomposition.

The invention discloses a dynamic representation learning method and device of a multiple heterogeneous information network, and relates to the field of graph representation learning. The method comprises the steps that at the t moment, a target network is decomposed into a plurality of sub-graphs according to the types of edges, and one sub-graph comprises one type of edges; for each sub-graph, performing causal multiple contrast learning to obtain a first learning result of the sub-graph, and further obtaining a first learning result of the target network; for each sub-graph, performing multiple sub-graph element learning to obtain a second learning result of the sub-graph, further obtaining a second learning result of the target network, and fusing the second learning result of the target network with the first learning result to obtain a representation learning result of the target network at the t moment; and repeating the steps at the next moment, updating the representation learning result of the target network until the target moment is reached, and obtaining a final representation learning result. According to the method, effective representation learning of the dynamic multiple heterogeneous information network can be realized.

The invention relates to the technical field of power grid risk assessment, and discloses a power grid risk assessment method based on integrated load flow calculation and topology analysis, and the method comprises the steps: collecting and preprocessing power grid operation data, and organizing the preprocessed data into a four-dimensional high-order tensor structure; constructing a power grid topological relation tensor network; constructing an optimization problem through an L1 norm form, and solving the optimization problem by using an augmented Lagrangian multiplier method; the complex coupling relation among time, space and parameters is revealed by analyzing the relation among different dimension factor matrixes; performing risk assessment on different granularity levels under a multi-scale tensor analysis framework, and coordinating analysis results of each level to form comprehensive assessment; designing a tensor completion estimation algorithm, constructing an optimization model by using low-rank hypothesis, and solving a complete data set through a tensor low-rank decomposition method; according to the method, the high-order tensor representation and tensor network decomposition technology is adopted, so that the dimension and complexity of data are effectively reduced, and the calculation complexity is reduced.

The application discloses a kind of roadbed pavement test and data acquisition system under mobile load, comprising: collecting high-precision road surface roughness data, and inhibiting noise by space-time convolution variation auto-encoding network decomposition feature.Based on normal stiffness and morphological gradient, dynamic grid division is used to adjust sensor density, and multi-modal data is fused using Bayesian weight. Adaptive signal details are extracted using multi-scale transformation, and filter is designed to enhance damage signal and optimize structural entropy. A multi-field coupled grid control equation is constructed to realize dynamic topology optimization reconstruction. A damage evolution state space is established, and a prediction model is trained through a multi-stage reinforcement learning framework combined with experience replay. A digital twin mapping model is constructed, integrating Bayesian optimization and reinforcement learning to establish a virtual-real feedback channel for dynamic decision optimization. The application not only provides innovative improvements to traditional test methods and data acquisition systems, but also accurately simulates and analyzes complex nonlinear dynamic behavior.

Techniques for using a network to decompose an HRTF data set to generate approximation data and to render an audio signal using the approximation data are disclosed herein. An input HRTF data set is fed into the network, which then generates approximation data that includes mixing channel gains, FIR filter coefficients, and basis filter shapes. This approximation data controls various components in the network. When the input HRTF data set is fed as input into the network, then the output of the network is an output approximated HRTF data set. The network iteratively fine tunes the approximation data until the output approximated HRTF data set sufficiently matches the input HRTF data set. After the approximation data is sufficiently tuned, the approximation data is later used by the network to render an audio signal.

The invention provides a method for realizing millimeter wave MIMO (Multiple Input Multiple Output) system channel estimation based on random tensor network decomposition, which comprises the following steps: k subcarriers and a transmitting signal on a tth time frame are acquired, a transmitting antenna transmits the transmitting signal to a receiving antenna, and the transmitting signal is expressed as xk (t) = sk (t) ck (t), # imgabs0 #, sk (t) represents an orthogonal pilot signal on the kth subcarrier, Ck (t) represents a precoding matrix on the kth subcarrier, and k subcarriers are represented as x (t) = [x1 (t),..., xK (t)] T on a time frame; through the random grid tensor decomposition method, complex signal processing in the millimeter wave large-scale MIMO antenna array is solved, accurate estimation of channel state information and wireless data transmission with cooperation of multiple communication modes are achieved, the flexibility and reliability of the system are improved, and the requirements of different application scenes are met. The adaptive network adaptation and service are provided for the user, the user experience is improved, and the high-quality communication service can be obtained in different network environments.

The invention relates to the technical field of unmanned aerial vehicle cluster control, and discloses a super network-based unmanned aerial vehicle cluster key point calculation method, which comprises the following steps of S1, constructing a super network of an unmanned aerial vehicle cluster; s2, constructing an unmanned aerial vehicle cluster sub-network based on the super network; s3, calculating a sub-network weight; and S4, calculating a key point deviation degree based on a key point deviation degree algorithm. According to the method, by introducing super-network modeling, function-driven sub-network decomposition and a multi-dimensional threat quantification mechanism, key point analysis scientificity and actual combat effectiveness are remarkably improved, unmanned aerial vehicle cluster task characteristics can be comprehensively considered, the relation between nodes in a network attack system is analyzed and explored, the key points of an unmanned aerial vehicle cluster are identified, and the risk of the unmanned aerial vehicle cluster is reduced. And the speed is higher, and the expandability is higher.

The invention discloses a sound velocity field reconstruction method based on full-connection tensor network decomposition and regularization, belongs to the technical field of ocean three-dimensional sound velocity field reconstruction, is used for ocean three-dimensional sound velocity field reconstruction, and comprises the following steps: obtaining a sparse three-dimensional sound velocity field observation value, inputting FCTN-T for processing, and obtaining a reconstructed ocean three-dimensional sound velocity field; a self-adaptive rank increasing strategy is added into the FCTN-T, and the self-adaptive rank increasing strategy comprises the steps of setting an initial value of a rank of an FCTN model, calculating a maximum rank of the FCTN model, iteratively updating parameters and obtaining a reconstructed ocean three-dimensional sound velocity field after a convergence condition is reached, so that the reconstruction efficiency is greatly improved. According to the method, the spatial correlation and the local smoothness characteristic of the three-dimensional sound velocity field are fully utilized, the best reconstruction precision is obtained in the aspects of overall and small-scale details, and particularly, a good reconstruction effect is achieved in a sparse sampling region; and by adopting a rank increasing strategy, the model is greatly improved in the aspect of operation efficiency, namely, the reconstruction precision and the reconstruction efficiency are considered at the same time.

The application discloses a heterogeneous network topology mapping method for power grid offline and online data, and has the characteristics that: according to the electrical characteristics of primary equipment of a power grid and inherent characteristics of the power grid, the primary equipment topology network of the power grid is decomposed into two-level networks; the first-level network is composed of power stations and lines, and forms a power station-level topology network; the topology network is homologous and isomorphic, and a subnet isomorphic algorithm of graph theory is applied to topology mapping; the second-level network is composed of primary equipment of each power station, and offline data and online data form topology networks according to the connection relationship between the primary equipment and the primary equipment; the first-level network and the second-level network expand the topology network of the power stations and the lines of the first-level network to the topology network of the primary equipment through the connection relationship of busbars and lines, realize the whole-network topology mapping of the offline and online data, and realize the primary equipment "node corresponding" mapping of the offline data and the online data to the whole-network "topology network" mapping.

The invention discloses a method for automatically generating a beam and a plate based on column structure decomposition and axis net connection. According to the method, geometric information of columns is extracted from a two-dimensional structure planar graph, through the processes of axis net decomposition, adjacent coupling beam pairing, degree sealing and surface domain tracking plate forming, a beam net frame is automatically generated according to the positions of the columns, a floor slab surface is finally generated, and therefore the complete topological relation among the columns, the beams and the slabs is established. According to the method, generation of wrong beam segments or wrong connections can be effectively avoided through a pairing strategy of pseudo interval filtering and outer contour priority. Therefore, the stable automatic modeling process of automatically generating the beam from the column and automatically generating the plate from the beam is realized. The method is used in a two-dimensional building structure plane graph, the arrangement of beams is automatically generated according to the positions and shapes of columns, and the floor outline is formed through closing. According to the method, manual participation can be reduced, manual omissions and errors are avoided, the modeling efficiency is greatly improved, and an accurate data basis is provided for engineering quantity calculation and cost analysis.

A decoding method is disclosed. Coefficients for a first set of layers of an (Implicit Neural Representation) network decomposed into a first set of layers and a second set of layers and parameters of the second set of layers are decoded. Parameters of the first set of layers are 5 determined as a linear combination of basis functions weighted by the coefficients, the basis functions being basis functions of a multiscale dictionary. An image or 3D scene is reconstructed based on the INR network using the parameters of the first set of layers and of the second set of layers.