15 results about "Sparse structure" patented technology

This invention discloses a monostatic symmetric nested array MIMO system and a near-field positioning method, belonging to the field of monostatic MIMO near-field positioning technology. This invention can effectively improve the positioning performance during near-field estimation. The monostatic symmetric nested array MIMO system proposed in this invention has a symmetrically distributed transmit and receive arrays. Compared to other symmetric MIMO systems, its differential array is a positionally continuous array with more degrees of freedom. With the same physical array aperture, fewer physical array elements are actually used, reducing costs. The near-field positioning method proposed in this invention calculates the fourth-order statistics of the signal and quantizes them to obtain a virtual differential array of the differential array, achieving accurate near-field positioning of sparse MIMO structure arrays. Furthermore, it can be extended to more sparse MIMO symmetric systems.

This invention provides a flexible production line control method and system based on a large vertical domain model. The method includes: constructing a pruning index combining temporal stability and spatial similarity for the original large vertical domain model to evaluate the redundancy of the activation tensors in the intermediate layers, and performing structured pruning accordingly to obtain a sparse structure model; performing mixed-precision quantization on the pruned model: allocating quantization bits according to the sensitivity of each network layer weight to task loss, and calculating scaling factors and zeros based on the range of non-zero weights to represent sparse weights; calculating the output deviation between the quantized pruned model and the original model under the same input, and training a feedforward network with the former's output as input to fit and compensate for the output deviation; combining the quantized pruned model and the feedforward network to form a composite control model; after deployment, adding the output command of the quantized pruned model to the compensation vector during inference to generate control commands.

The application provides a high-dimensional time-varying signal space completion method and device based on a sparse attention mechanism, equipment and a storage medium, and relates to the technical field of high-dimensional information. The method divides original data into multiple sparse structured sub-blocks through adaptive blocking; performs low-rank-sparse structure decomposition on each sub-block, extracts dominant low-rank structure features and sparse disturbance features, and fuses to obtain sub-block-level sparse structure representation; a sparse attention mechanism with triple constraints of time sequence, missing mode and structure recoverability is constructed, a dependency relationship is established only between key block pairs, and a space-missing joint dependency representation is output; through multiple rounds of iterative feedback, the updated matrix is re-input into the foregoing operation until the error converges, and a completely completed high-dimensional matrix is output. The application can effectively suppress redundant correlation, enhance dependency modeling of the completable area, and significantly improve the completion accuracy and robustness of high-dimensional time-varying data in a complex environment.

The present application discloses methods, systems and computer program products for sparse quantization of a model. The method comprises: generating a fine-tuning dataset based on data associated with a target task of a pre-trained model; performing multiple rounds of progressive sparsification on the pre-trained model; and performing multiple rounds of progressive quantization on the model after the multiple rounds of progressive sparsification. Each round of progressive sparsification comprises: performing a sparse operation; performing hierarchical fine-tuning on the model resulting from the sparse operation using the fine-tuning dataset, and imposing a sparsity structure constraint during the hierarchical fine-tuning. Each round of progressive quantization comprises: performing a quantization operation; performing hierarchical fine-tuning on the model resulting from the quantization operation using the fine-tuning dataset, and imposing both a sparsity structure constraint and a quantization structure constraint during the hierarchical fine-tuning. The method achieves effective maintenance of the performance of the model on the target task while significantly reducing the model storage footprint and computational complexity.

This invention discloses a terahertz multi-component detection method for hydrogen-fired engine exhaust gas based on sparse inversion. The method includes: selecting a predetermined number of discrete frequency points in the terahertz band as observation frequency points, acquiring discrete spectral data, and preprocessing it to obtain an observation data vector; parameterizing the absorption characteristics of multiple gas components in the exhaust gas based on the observation data vector to construct a feature dictionary; utilizing the sparse structure where the number of gas components in the actual exhaust gas is less than the size of the feature dictionary, solving the parameter vector by introducing sparse constraints, identifying the gas type based on the feature atoms corresponding to the significantly non-zero parameter components in the parameter vector, and estimating the concentration of the gas component based on the sum of the significantly non-zero parameter components belonging to the same gas component. This invention, through sparse frequency sampling and sparse inversion, achieves rapid identification and concentration inversion of multiple gas components in hydrogen-fired engine exhaust gas without relying on the complete terahertz spectrum, significantly reducing measurement and computational overhead.

This invention provides a short text clustering and fuzzy recognition algorithm for large-scale online subgraph sampling, comprising the following steps: Step S1, extraction and preprocessing of training samples; Step S2, construction of the neural network; Step S3, overall clustering prediction; Step S4, fuzzy sample recognition; Step S5, retraining of the neural network. This invention combines short text clustering with a large language model, which not only improves clustering accuracy but also enables the handling of clustering tasks with different themes and classification requirements, significantly reducing the manual cost of data annotation. Furthermore, this invention can annotate fuzzy samples for classification, using K-nearest neighbors combined with minimum spanning trees to assist subgraph sampling in the selection scheme. This utilizes sparse structure to reduce computational costs and exposes the fluctuations of boundary samples through spectral clustering, providing a more comprehensive perspective for fuzzy sample selection and improving the accuracy and interpretability of the clustering results.

This invention discloses an adaptive granularity management system for a large language model KV cache based on access co-occurrence awareness, relating to the field of large language model technology. It includes a pattern feature extraction module, a head group clustering construction module, a flexible boundary generation module, a semantic weight calculation module, a hierarchical compression storage module, and a data merging and scheduling module. The pattern feature extraction module performs sparse structure analysis on the attention weight matrix of the large language model to extract attention pattern fingerprint data. The head group clustering construction module receives the attention pattern fingerprint data, aggregates multiple attention heads with similar pattern features into head tuples, and generates a parameterized co-occurrence template. The flexible boundary generation module obtains the current inference sequence length and, combined with the parameterized co-occurrence template, converts the relative position parameters within the template into a dynamic mapping table through proportional scaling. The semantic weight calculation module calculates the range of co-occurrence groups based on the dynamic mapping table.

This invention provides a UPA near-field channel estimation method based on two-dimensional block sparsity, comprising: representing the UPA near-field channel matrix as the sum of the outer products of the horizontal and vertical ULA near-field steering vectors; constructing improved DFT dictionaries in the horizontal and vertical directions respectively; representing the channel matrix as a total coefficient matrix under these dictionaries, which presents a two-dimensional block sparse structure determined by the outer product of the horizontal and vertical block sparse representation vectors, thereby transforming channel estimation into a two-dimensional block sparse signal recovery problem; and solving the problem using the 2D-PCSBL algorithm, in which the accuracy parameter of each sparse coefficient in the prior distribution is determined by the weighted sum of its own hyperparameter and the hyperparameters of its two-dimensional neighbors, to capture the sparse coupling characteristics of the UPA near-field channel in the horizontal and vertical dimensions. This invention achieves high-precision channel estimation with fewer pilots and a lower signal-to-noise ratio, and its computational complexity is comparable to that of the one-dimensional block sparse method.

The application provides a message passing based multi-task clustering sparse reconstruction method. The method uses the joint clustering sparse structure characteristics of sparse signals among different tasks, so that better sparse reconstruction performance is obtained under the condition of fewer observation samples. Specifically, the sparse structure characteristics of the clustering sparse signal are described by using a Markov Spike and Slab prior, a generalized approximate message passing algorithm is introduced to iteratively approximate the posterior mean of each unknown variable, and an expectation-maximization method is used to iteratively update the unknown parameters. Compared with the traditional single-task Bayesian compressive sensing algorithm, the method has a significant performance improvement under the condition of fewer observation samples.

The application discloses a kind of diffusion model image generation methods and medium based on gamma noise.The method obtains pure gamma noise tensor and generates initial state tensor as physical base by centralization processing;For each current time step of the preset continuous discrete time grid, generate effective diffusion coefficient, and utilize the score network trained in advance to execute forward action to current state tensor, generate edge score approximation;Based on probability flow ordinary differential equation, numerical integration state update calculation is carried out using the above parameters, to generate previous time step state tensor;In turn cycle until time is zero, execute anti-normalization and numerical truncation operation to output target generated image to final state tensor.The application breaks through the limitation of traditional Gaussian hypothesis, and improves the generation quality and efficiency of complex skewness and sparse structure image.

An optimization method for sparse matrix-vector multiplication of multiple matrices with the same structure is applied to the iterative solution process of harmonic balance methods in radio frequency circuit simulation. The method includes: obtaining a first sparse matrix and a second sparse matrix of the circuit, wherein the first and second sparse matrices have the same structure and are used to characterize the nonlinear conductance and nonlinear capacitance characteristics of the circuit, respectively; in response to performing matrix-vector multiplication, in one traversal, simultaneously accessing the non-zero elements located at the same position in both the first and second sparse matrices, and performing multiplication and accumulation operations with the corresponding input vector elements to obtain the output vectors of the corresponding first and second sparse matrices; wherein the input vector is the harmonic component vector of the circuit node voltage. This invention, based on the combined access of multiple matrices with the same sparse structure, significantly reduces loop control overhead, thereby improving the overall efficiency of matrix-vector multiplication.

This invention discloses a beam-domain channel estimation method for spatially non-stationary large-scale MIMO, comprising the following steps: First, a beam-domain channel model for spatially non-stationary large-scale MIMO is constructed using the visible region; then, based on the sparsity of the beam-domain channel and the influence of power leakage, the beam-domain channel estimation problem is transformed into a sparse channel reconstruction problem; next, based on the sparse structure of the cross-blocks of the beam-domain channel and the power ratio threshold, a sparsity-adaptive matched pursuit scheme based on the beam-domain structure is proposed; finally, simulation results verify that the proposed scheme has lower pilot overhead and higher accuracy and effectiveness compared with traditional schemes. The beam-domain channel estimation method for spatially non-stationary large-scale MIMO provided by this invention can be effectively applied to channel estimation with non-stationary characteristics, and has significant advantages in estimation accuracy and complexity.

This invention discloses an adaptive coverage method for swarm robots based on sparse large-scale optimization, comprising: constructing an environmental situation tensor and risk map using multi-source sensor data; dynamically generating adaptive behavior weights using an attention mechanism; calculating anisotropic behavior forces and synthesizing motion commands based on the weights and risk map; constructing and dynamically updating an importance flow graph representing the criticality of parameters by analyzing historical parameter data; performing sparse pruning and directed evolution optimization on swarm control parameters guided by the importance flow graph; and integrating anti-interference processing and fault tolerance mechanisms. This invention achieves context-aware dynamic adjustment of the interaction intensity of swarm robots in complex environments, improves the efficiency of large-scale parameter optimization by utilizing the sparse structure of parameters, and ensures the overall robustness and coverage efficiency of the system.

The application discloses a network anomaly tracing method and device based on a multi-modal causal graph and a large language model reasoning and a medium, the method constructs a multi-modal heterogeneous causal graph fusing index, call chain and log core clues, and designs a cross-modal reasoning mechanism for a large model, realizing comprehensive analysis of the root cause of the anomaly; for multi-modal observable data, the causal relationship of the time series index data is modeled by using sparse structure learning and independence test, a global causal graph is constructed in combination with link connectivity probability, and a semantic perception strategy is used to extract a log sequence semantic abstract of the associated node, so that the multi-modal observable data is uniformly organized in the multi-modal heterogeneous causal graph; the large language model reasons the root cause of the anomaly on the multi-modal heterogeneous causal graph, and finally generates a tracing result response text with professionalism and interpretability, thereby providing a new solution for multi-modal information fusion and anomaly tracing of a distributed complex network.