27results about How to "Avoid computational overhead" patented technology

The invention provides a channel prediction method and system based on a multi-cycle feature convolutional network. According to the method, context sensing weighted input and a dynamic multi-model pool mechanism are introduced. A terminal device obtains environmental context features in real time, dynamically calculates the weight of each time step in a historical channel state information sequence based on the features, and generates a weighted optimized input sequence. Meanwhile, the network equipment pre-trains and maintains a plurality of multi-period feature convolutional network models specific to different environments, and the network can effectively capture the complex time-varying rules in the period and during the weeks in the channel, so that the accuracy of channel state information prediction is improved. And the terminal equipment immediately selects and activates the optimal sub-model in the model pool for prediction according to the real-time context features, and locally performs increment fine adjustment. According to the invention, through management of the data input end and rapid and adaptive switching of the prediction model end, the precision of channel prediction and the adaptive capacity to environmental sudden change are effectively improved.

The application relates to the technical field of chip design, in particular to an AI-based DDR5 memory interface chip adaptive method and system.The method comprises the following steps: obtaining a configuration instruction stream, extracting current and historical configuration instructions, and predicting target pins, switching time and target modes.Before the switching time, target electrical parameters are calculated according to pin electrical characteristics, historical switching records and channel load characteristics.During the switching, the target parameters are loaded and the temporary impedance matching network is switched in.After the switching, a fine-tuning increment is generated according to the deviation of the signal quality and the threshold value, and the updated electrical parameters are obtained by superposition, and the matching network is disconnected after the updated electrical parameters are written back to the register.According to the timing error and the fine-tuning increment generated by multiple switching, the weight parameters are updated and deployed back to the chip.Through AI prediction and the dynamic switching and fine-tuning of the temporary impedance matching network, seamless, stable and low-delay mode switching of the memory interface chip in the reconfigurable computing system is realized.

This invention relates to a method for patent retrieval and technology monitoring of multi-source heterogeneous data. It constructs a multimodal semantic representation including text, images, and time-series fragments, and generates a joint temporal representation vector by combining attention perturbation entropy and term evolution distance. A unified semantic space is established based on historical technology anchors, dynamically monitoring and quantifying semantic shifts to achieve real-time determination of drift trends. When substantial semantic drift occurs, local semantic correction is performed using gated contrastive learning and an anchor-aware adaptation layer to guide the orderly alignment of newly added modal features, ensuring the continuation and expansion of the original semantic topology. Simultaneously, an anchor decay and candidate new anchor generation mechanism is designed, combining cross-validation from patent citations, industry reports, and policy documents to achieve incremental expansion of the semantically stable anchor set. This scheme effectively improves the fusion accuracy, real-time adaptability, and dynamic topology update capability of multimodal patent data.

The present application relates to the field of simulation technology, specifically to a device cluster job simulation planning method based on multi-agent cooperation, which constructs a land grid model and a yield density upper and lower bound channel, determines the capacity critical threshold interval by forward cumulative scanning, generates a target point based on the interval midpoint and passability constraints, and realizes cooperative device path guidance. The target point offset gating mechanism is introduced to control path re-planning triggering, improve path stability and responsiveness. The interval boundary is dynamically modified during the job process by combining the measured yield, and the final meeting point is locked when the interval length and meeting distance conditions are met. After the simulation is completed, the attribution analysis is performed by locking the deviation and cumulative displacement indicators, the row sequence is adjusted and iteratively re-simulated and optimized by combining the row accessibility, and the optimal job scheme is output. The method can improve the efficiency and simulation accuracy of device cooperative operation in complex environments.

The application provides a power grid fault traveling wave positioning method based on a fusion optimization algorithm, comprising: establishing a dynamic virtual fault framework model of a power grid, randomly generating a plurality of virtual fault particles containing position coordinates, and constructing an initial virtual fault particle cluster; using a shortest path algorithm to calculate the topological path length of each virtual fault particle in the virtual fault particle cluster to all measurement points in the power grid, and then calculating a virtual fault space-time feature vector; setting the traveling wave arrival time when a real fault occurs as a target clustering centroid in a feature space, and calculating a feature matching fitness; constructing a non-convex optimization problem with the optimal feature matching fitness as a target, solving the problem using a fusion optimization algorithm, and thus obtaining a fault positioning result and a traveling wave velocity, so as to realize traveling wave positioning. The application can solve the problems of local optimum, slow convergence speed and high calculation cost in the prior art.

The invention provides a false data injection attack method for a multi-elastic-joint robot system, and belongs to the technical field of multi-agent network attacks based on computer data processing. Firstly, a multi-elastic joint robot nonlinear dynamic mathematical model is constructed and converted into a high-order all-drive quasi-linear multi-robot system dynamic model based on an all-drive system theory; and then designing a distributed consistency controller containing a linearization item and a neighbor error related item, establishing closed-loop global dynamics and deducing a state analytical solution in combination with equivalent conversion of the model. Designing a false data injection attack strategy, and constructing an attacked system dynamic model; a closed-loop terminal state error in the presence or absence of attacks is taken as a target function, an attack optimization problem is converted into a combinatorial optimization problem, a greedy algorithm is designed by applying a sub-module optimization theory, and a suboptimal solution in polynomial time is obtained and is taken as an optimal scheme of false data injection attacks of the system. According to the method, the applicability of the complex nonlinear multi-agent system is improved, and the calculation complexity is reduced.

The application provides an improved tabu search topology optimization method of a marine wind farm power collection system, which is based on the positions of wind turbines to intelligently preliminarily group and sort, to form an initial connection scheme, to connect all the wind turbines in each group with the minimum cost, to calculate the group cable cost, to check whether the lines cross, to impose a penalty in the total cost if the lines cross, to ensure that the scheme is feasible, to compare the current scheme with the historical best, to update the record and intelligently exchange the wind turbine groups to explore a new scheme, to repeat the optimization process until the number of iterations is satisfied, and to finally output a globally optimal topology scheme. The application has the advantages of better optimization effect, significantly improved economy, higher calculation efficiency, stronger engineering practicability, better algorithm robustness, better scheme reliability, high design automation degree and strong auxiliary decision-making capacity.

The present application relates to the technical field of model optimization, and is a neural network training and optimization system for medical knowledge, specifically comprising: boundary feature collection on medical knowledge training samples to obtain boundary indexes reflecting the relationship between conclusions and condition distribution; forming a condition decoupling representation according to the boundary indexes, which is used to describe the separation trend between conclusions and applicable conditions; using the condition decoupling representation to correct the evaluation of model training intermediate quantity, obtaining sample-level boundary loss indexes, and aggregating to obtain knowledge cluster-level boundary loss indexes; jointly mapping the knowledge cluster-level boundary loss indexes and external disturbances to obtain boundary expansion probability and enclose the knowledge cluster to be reviewed according to the boundary expansion probability; using probe samples to perform targeted verification on model output, and tracing abnormal output to knowledge anchor points to be corrected to perform differential optimization processing. The present application solves the problem of condition boundary loss of neural network models in the prior art.

The application relates to an LED display screen display non-uniformity adjusting method and system based on image processing, which comprises the following steps: converting the obtained real-time multi-source physical state parameters into a two-dimensional pixel-level feature map through spatial interpolation mapping; obtaining a transient environmental deviation matrix, combining a cumulative lighting duration map to obtain a long-acting aging deviation matrix; weighting and fusing the two to obtain an overall deviation matrix, and generating a target update instruction set according to the comparison result of the regional gray variance of the overall deviation matrix and a preset gradient threshold; in response to the instruction set, superimposing a preset static reference correction matrix and the overall deviation matrix to generate a preliminary dynamic correction matrix, and performing boundary smoothing processing on the preliminary dynamic correction matrix through a spatial edge-preserving filtering algorithm to obtain an adaptive target correction coefficient matrix; finally, mapping and converting a video source data frame and the matrix point by point into an actual driving signal rendering output. The application can improve the accuracy of LED display screen display non-uniformity adjustment.

This invention provides a method and apparatus for decision-making and execution in a creative workflow based on an AI agent. The method includes: normalizing natural language instructions to obtain normalized text; inputting the normalized text into a rule engine to generate a first modality and its first confidence level; if the first confidence level is greater than or equal to a first threshold, determining the first modality as the target modality; otherwise, inputting the normalized text into a pre-trained modality classification model for inference to generate a second modality and its second confidence level; if the second confidence level is greater than or equal to a second threshold, determining the second modality as the target modality; otherwise, triggering a user fallback confirmation process to receive user confirmation of the target modality; and forwarding the natural language instructions to the work creation engine corresponding to the target modality to generate the target work. This invention achieves automated and intelligent determination of creative modalities, simplifies user operation processes, and improves user interaction experience and the efficiency of creative workflow decision-making and execution.

The application provides an enhanced generation method based on topology-aware graph coding and adaptive subgraph retrieval, including multi-granularity semantic-aware segmentation and knowledge graph construction, structure feature matrix construction and explicit topology position coding generation, constrained subgraph diffusion and dynamic pruning based on semantic-topology joint scoring, and graph-text consistency loss constraint generation based on attention matrix structured alignment; in the coding layer, explicit topology position coding is used to avoid the overhead and excessive smoothing caused by online GNN aggregation; in the retrieval layer, connected subgraphs instead of fragmented nodes are used as enhanced context; in the alignment layer, the internal attention matrix of the large language model is directly constrained to be consistent with the subgraph adjacency matrix, the structured guidance is realized from the reasoning mechanism level, so that the logical illusion is inhibited and the multi-hop reasoning accuracy is improved.

This invention relates to a multipath adaptive full-focus ultrasound imaging method for corner areas of composite materials. The method includes: acquiring full-matrix ultrasound data of the corner area of ​​the composite material; constructing a quasi-P-wave group velocity sound velocity model based on the characteristic parameters of the composite material; constructing a surface-layered discrete model with a local coordinate system, and abstracting it into a directed graph G based on the quasi-P-wave group velocity sound velocity model; searching for a set of candidate propagation paths from array elements to arbitrary nodes in the directed graph G, and constructing multiple full-focus imaging TFM images based on the candidate propagation paths and the full-matrix ultrasound data; calculating path score factors based on the multiple full-focus imaging TFM images, and performing pixel-by-pixel weighted fusion of the multiple full-focus imaging TFM images based on the path score factors to obtain a fused imaging result. Compared with existing technologies, this invention has advantages such as strong adaptability and improved focusing stability and imaging reliability in corner areas.

This invention discloses a privacy-preserving vector database query method, comprising: each server obtaining a query vector secret share and performing a linear transformation to obtain a transformed query vector secret share; sending the locally held transformed vector data secret share and the transformed query vector secret share to semi-trusted hardware; the semi-trusted hardware performing a nearest neighbor search to obtain the nearest neighbor vector and its corresponding index; sharing the nearest neighbor vector and the one-hot vector corresponding to the index with each server; each server performing a dot product calculation between the one-hot vector secret share and the locally held label secret share to obtain a query result label secret share; performing an inverse linear transformation on the nearest neighbor vector secret share to obtain a query result vector secret share; and sending the query result label secret share and the query result vector secret share to the user terminal; the user terminal reconstructing the query result vector and the query result label.

The present application relates to the technical field of underwater acoustic signal detection, and particularly relates to a weak underwater acoustic signal adaptive detection method based on sound field complexity. The method comprises the following steps: underwater acoustic signal collection and preprocessing; multi-dimensional feature analysis is performed on the preprocessed underwater acoustic signal, and multi-dimensional features reflecting the current sound field environment complexity are extracted; the multi-dimensional feature parameters extracted in step S2 are fused to generate a comprehensive complexity score, the complexity level of the current underwater acoustic environment is evaluated, and adaptive detection algorithm decision is made according to the sound field complexity level obtained after evaluation; a specific underwater acoustic signal detection task is executed, weak underwater acoustic signals are detected and processed, and corresponding detection algorithms are executed according to the sound field complexity level; and the detection result is output. The method can dynamically select a suitable detection algorithm according to the change of the underwater acoustic environment, thereby improving the detection stability and reliability of weak underwater acoustic signals in a complex environment.

A method and system for image similarity recognition based on multi-feature fusion, relating to the field of image processing technology, is disclosed. The method includes: calculating the hash features of an input image and generating hash similarity and hash confidence values; when the hash confidence value is less than a hash confidence reference value, performing local feature extraction on the input image to generate local similarity and local confidence values; when the local confidence value is less than the local confidence reference value, performing deep learning feature extraction on the input image to obtain deep features and deep confidence values; determining the weight coefficients of the hash features, local features, and deep features based on the hash confidence value, local confidence value, and deep confidence value, respectively; performing feature fusion to generate a fused feature vector; and calculating the final similarity between the input image and the image to be compared based on the fused feature vector. Implementing this application can improve processing efficiency while ensuring the accuracy of image similarity recognition.

The invention discloses an SATA (Serial Advanced Technology Attachment) bus anomaly detection method, system and device for local cache of an edge gateway, and relates to the technical field of edge computing and bus detection. According to the SATA bus anomaly detection method for the local cache of the edge gateway, the time sequence offset characteristic in the SATA bus link establishment stage and the statistical distribution characteristic in the data transmission stage are obtained, the health degree is calculated in a fusion mode and filtered, the protection priority is determined in combination with the data importance level and the service load, the cache writing strategy is matched, and the SATA bus anomaly detection method for the local cache of the edge gateway is achieved. According to the method, the characteristics of the SATA bus link establishment stage and the data transmission stage are obtained, a related model is constructed, the bus health degree is calculated in a fusion mode, and the adaptive cache protection strategy is matched, so that the safety of the SATA bus link establishment and data transmission stage is improved, and the safety of the SATA bus link establishment and data transmission stage is improved. Accurate detection of bus abnormity is realized, and security of local cache data of the edge gateway is guaranteed.

The invention provides a sensitive attribute forgetting method and device for a recommendation system. An embedding layer of the recommendation system comprises a user embedding matrix and an article embedding matrix; the method comprises the following steps: performing matrix decomposition on an article embedding matrix to obtain an independent space orthogonal basis; performing feature decoupling on the user representation vector to obtain a correlation vector and an irrelevant vector; the irrelevant vector is represented as a linear combination of a to-be-optimized parameter and an irrelevant space orthogonal basis; constructing user representation distribution and sensitive attribute distribution, and selecting the optimal bandwidth of the HSIC core; constructing a forgetting loss function used for measuring the dependency between the user representation distribution and the sensitive attribute distribution, and replacing a user representation vector in the forgetting loss function with a combination of a correlation vector and an irrelevant vector to obtain a target function; and under the condition of keeping the correlation vector unchanged, taking a minimization objective function as an objective, obtaining a final user representation vector through iterative optimization, and taking the final user representation vector as a result after the sensitive attribute is forgotten.

This invention relates to the field of power communication network technology and discloses a method for automatic discovery and visualization of power communication network topology. The method includes receiving a set of physical device nodes and physical links, assigning spatial hash codes with spatial hierarchy prefixes; appending links to an incremental graph buffer when a link undergoes a state change, extracting prefixes to determine cross-domain links, calculating damping states based on a time decay mechanism, and locking stable contribution values ​​accordingly; triggering merging when buffer records reach a threshold, calculating the contribution value difference and accumulating it into the upper-layer aggregated edge weights, updating degradation labels based on damping states and persisting them to a persistent storage module; receiving topology retrieval requests, blocking or opening buffer access based on the comparison result between the unmerged event hierarchy depth and the viewport hierarchy depth, and returning data. By blocking the propagation of high-frequency changes in lower-level links to the upper layers, repeated front-end rendering is avoided, improving the efficiency of network topology data processing.

This invention belongs to the field of load forecasting, and specifically relates to an automatic optimization method and system for electricity spot trading strategies. It includes: a data acquisition and processing module for acquiring and preprocessing multi-source time-series data; an electricity price forecasting module for inputting high-dimensional feature vectors into an electricity price forecasting model and outputting high-dimensional electricity price forecast vectors for future scheduling cycles; a strategy generation module for inputting high-dimensional electricity price forecast vectors into a strategy generation model and outputting a trading strategy parameter set; and an adaptive calibration module for online calibration of the electricity price forecasting model according to a preset calibration strategy. The preset calibration strategy involves real-time monitoring of the moving average of the high-dimensional electricity price forecasting error and the statistical drift index of the data distribution. When the error threshold is exceeded or the drift exceeds a preset confidence interval, an incremental parameter update mechanism is triggered. This invention solves the problem in existing technologies where static forecasting models cannot adapt to the non-stationary environment of the power system, leading to a decrease in forecast accuracy and consequently, strategy failure.

The application provides an unmanned aerial vehicle (UAV) insulator string defect detection method, which comprises the following steps: obtaining a target image of an insulator string photographed by the UAV, and constructing a defect detection model based on an improved YOLOv5 convolution layer; performing defect detection on the target image by using the defect detection model to obtain a defect detection result of the insulator string; wherein the defect detection model comprises a grouping dimension increasing module, a multi-scale attention module and a double-path collaborative detection head, which can reduce the model parameter quantity while improving the high-dimensional feature extraction capability of the insulator string, realize rapid detection and positioning of insulator cracks, contamination and other defects, effectively suppress complex background interference through the multi-scale attention module, and enhance the robustness of defect detection; and inference is performed through the lightweight double-path collaborative detection head, which can enhance the feature diversity while avoiding additional calculation overhead in the inference stage, thereby improving the end-to-end defect detection capability of the insulator string of the UAV.

This invention relates to the field of blockchain technology and provides a method, apparatus, electronic device, and storage medium for tracing electricity consumption data. The implementation scheme is as follows: A preset electricity traceability model is invoked to standardize and model the electricity consumption data of each user, obtaining various triplet data sets, which are then stored on the blockchain; a composite index structure is constructed based on the first field of each triplet data set, wherein the composite index structure includes a traceability positioning table and a multi-bucket index structure; responding to a query request from a target user, cross-block positioning is performed based on the second field in the query request and the traceability positioning table to obtain a cross-block traceability link; according to the second field and the multi-bucket index structure, intra-block data extraction is performed on each block in the cross-block traceability link to obtain the target traceability result; the target traceability result is then sent to the target user. This invention can improve traceability efficiency.

The application discloses a kind of based on the self-adapting data blood relationship construction method of double engine cooperation, belong to enterprise-level data governance platform technical field, including obtaining the heterogeneous data processing script to be handled and pre-processing, obtain structured intermediate representation, respectively to structured intermediate representation Static syntax analysis and semantic reasoning based on pre-training language model are executed, generate first blood relationship and second blood relationship, carry out conflict resolution and fusion, obtain fusion blood relationship, execute cause-effect logic check to fusion blood relationship, and modify the dependent relationship that does not pass the check, obtain target blood relationship, in response to monitoring that heterogeneous data processing script changes, determine the processing link affected, and only the structured intermediate representation corresponding to the processing link affected is reprocessed, realize the incremental adaptive update of blood relationship;The application improves the accuracy and adaptive capacity of blood construction.

This invention discloses a method for optimizing the bandgap of tandem solar cells based on limiting efficiency. First, an overall structural model of the tandem solar cell is established, and material parameters and incident spectrum boundary conditions are set. An optical solution model is established for the top cell region. Optical solutions are then performed on each sub-cell from top to bottom to obtain the photogenerated carrier generation rate distribution and transmission results for each layer. The incident spectrum is updated layer by layer based on the transmission results of the previous layer until the entire layer calculation is completed. The photogenerated carrier generation rate of each layer is used as electrical input for electrical analysis, and the J-V curve of current density-voltage for each sub-cell is obtained through electrical solution. The limiting efficiency is calculated based on the J-V curve to find the maximum power point. The optimal bandgap is determined based on the maximum power point and the limiting efficiency. By using optical-electrical multiphysics coupling modeling and introducing layer-by-layer spectral transfer, a photogenerated carrier generation rate and J-V characteristics that more closely resemble actual operating conditions are obtained, thereby selecting a more realistic optimal bandgap.

The application relates to a charging pile cooperative power supply control method and system, a terminal and a storage medium, and relates to the technical field of charging piles. The method comprises the following steps: acquiring target demand power of a target charging vehicle; acquiring default power supply power of a target charging pile; judging whether the default power supply power is less than the target demand power; if yes, determining a power deficiency amount according to the target demand power and the default power supply power; determining a compensating charging pile and adjustable power of the compensating charging pile based on the power deficiency amount; connecting a target cross-pile power supply path between the compensating charging pile and the target charging pile; controlling the compensating charging pile to deliver the adjustable power to the target charging pile along the target cross-pile power supply path; integrating the default power supply power and the adjustable power to obtain target output power; and charging the target charging vehicle based on the target output power. The application has the effects of improving the charging utilization efficiency of the charging pile and enabling multiple charging piles to cooperatively output to meet higher target voltage requirements.

The invention discloses a path planning method, system and device for an unmanned ship and a storage medium, and the method comprises the steps: 1, collecting the environmental parameters of an area in a pond, constructing a global map, and inputting the geographic coordinates of a task area; setting bait casting point distribution and adjusting navigation parameters; 2, acquiring a current position, delimiting a safe navigation boundary, and ensuring that subsequent movement does not exceed a preset range; 3, carrying out route calculation based on an algorithm fusing a dynamic weight mechanism and an adaptive cost function, and generating a global route plan; according to global path planning, an overall driving route from a starting point to a target point is generated based on known environment information, the local path module senses the surrounding environment in real time depending on a sensor carried by the unmanned ship, dynamic data such as the obstacle position, the water flow direction and the speed are obtained, obstacle avoidance and course correction are achieved through rapid re-planning or behavior decision making, and the unmanned ship is more intelligent. According to the mechanism, the calculation overhead caused by complete re-planning is avoided, and the problem that the local part falls into the local minimum is also prevented.

The application relates to a self-adjusting data security control method, device, equipment, storage medium and program product. The method comprises the following steps: acquiring an environmental behavior feature in current access data, and determining a data risk value of a current access behavior according to the environmental behavior feature; in the case that it is determined that there is risk according to the data risk value and a preset data risk threshold value, determining a target security level of the accessed data; the accessed data represents an object accessed by the current access behavior; and performing a matched control operation on the current access behavior according to the target security level. The method can guarantee data security while improving business access efficiency.

This application discloses a positioning and navigation method, apparatus, device, storage medium, and computer program product to solve the problem that existing positioning and navigation schemes cannot provide accurate real-time navigation guidance to users in complex environments when a preset map is lacking. The method includes: performing feature recognition on at least two perspective 2D images corresponding to a target object to determine a target region in the 2D images; performing semantic segmentation on the target region to obtain a semantic segmentation mask corresponding to the target region; generating a 3D model corresponding to the target object based on the semantic segmentation mask and the 2D images; determining the pose information of the device to be navigated within the 3D model based on environmental video and motion data collected by the device to be navigated; generating a navigation path based on the position information of the target region and the pose information; and sending the navigation path to the device to be navigated for navigation guidance.