11results about How to "Guaranteed sparsity" patented technology

The invention relates to a satellite navigation and positioning method and device, and belongs to the technical field of satellite navigation and positioning. Resolving the GNSS observation value based on a single-difference GNSS sliding window optimization model to obtain a floating point solution corresponding to a positioning result; the sparsity of a matrix in a graph optimization problem is ensured based on a single-difference GNSS sliding window optimization model; further, gross error detection is carried out on the GNSS observation value based on sliding window data, a test result is obtained, and the robustness of the algorithm in a complex environment is ensured; when the test result is that the test is passed, determining a single-difference ambiguity variable according to the floating point solution, and performing inter-satellite single-difference on the single-difference ambiguity variable to obtain floating point double-difference ambiguity; the floating point double-difference ambiguity is fixed based on a partial ambiguity fixing method, the satellite navigation positioning process of a GNSS fixed hold solution (Hold solution) under a graph optimization model is realized, and the precision and reliability of satellite navigation positioning are improved.

The invention discloses a jacket platform reliability analysis method based on data combination dimensionality reduction and polynomial chaos expansion, which comprises the following steps of: performing dimensionality reduction on a jacket platform high-dimensional input sample by adopting a combination dimensionality reduction method, determining a transformation matrix, mapping the high-dimensional input sample and a test sample to a low-dimensional space, and determining the reliability of the jacket platform; the method comprises the following steps: preparing for establishing a dimensionality reduction PCE model, establishing a PCE model in a transformed input space, adaptively selecting a polynomial highest order and a regularization parameter by adopting a sparrow search algorithm, reconstructing a jacket platform PCE model according to the adaptively selected optimal parameter, and calculating a failure probability and a reliability index through random sampling and structural reliability analysis. According to the method, high-dimensional sample data of the jacket platform can be mapped to a low-dimensional space, the modeling efficiency is improved, parameters are adaptively selected, the modeling precision is improved, and high-precision and high-efficiency jacket platform structure reliability analysis is realized.

This invention belongs to the field of sensors and signal processing, specifically relating to a wavelet kernel-scale sensitivity-guided denoising method for abrasive particle induced voltage signals. The method includes: acquiring abrasive particle induced voltage signals and performing harmonic cancellation to obtain a preprocessed signal; constructing a wavelet kernel function and calculating a kernel-scale guided spectrum using it and the preprocessed signal; constructing a sparse joint denoising model based on the kernel-scale guided spectrum; processing the sparse joint denoising model to obtain a convex optimization objective function; solving the convex optimization objective function using an adaptive step-size gradient descent method and an adaptive iterative shrinking threshold method to obtain a weight vector characterizing the distribution of abrasive particle characteristic signals; binarizing the weight vector characterizing the distribution of abrasive particle characteristic signals to obtain a feature indicator vector; performing a Hadamard product between the feature indicator vector and the preprocessed signal, followed by low-pass filtering to obtain a denoised signal. This invention can adaptively and non-destructively enhance and denoise abrasive particle characteristic signals under strong interference environments.

The application provides an OTFS fractional Doppler multi-target detection method based on group sparsity, and the implementation steps are as follows: initializing parameters; constructing and solving a group sparse signal recovery problem in two stages of coarse estimation and fine estimation; and obtaining a multi-target detection result through the two-stage solving result. The application constructs and solves a group sparse signal recovery problem in two stages of coarse estimation and fine estimation, calculates the detection result of the multi-target through the real number domain channel gain of the two-stage solving, utilizes the complex number characteristics of the real number domain channel gain, maintains the group sparsity and non-sparsity within the group of the time delay-Doppler domain channel gain, and avoids the defect that the amplitude of the non-target position is increased due to interference and noise. In the group sparse signal recovery, the channel gain of other channels is updated through iteration, and the signal components corresponding to other channel gains are subtracted, so that the interference between targets is gradually eliminated, and the detection precision is improved.

The invention discloses an incoherent authorization-free communication method, device and system for dual-mode information grouping, and belongs to the field of wireless communication, and the method comprises the following steps: incoherent information modulation: a signal transmitting end determines the information transmission mode of each group in a data frame, the position of an active information transmission block and the pilot frequency selected by the active block according to system configuration information and a bit to be transmitted, the information transmission mode comprises a high-low density transmission mode and a low-density transmission mode, and the number of active data blocks in groups in the high-density transmission mode is large; energy estimation: the receiving end estimates the energy of the transmitted signal by adopting a block coordinate descent algorithm based on pruning operation; active sending end detection: setting an adaptive quantity threshold value based on an energy estimation vector estimation result, and determining an active sending end according to an energy threshold value and the number of active data blocks; and information bit recovery: carrying out information bit recovery on the active sending end according to the estimation result of the energy estimation vector and the configuration information. According to the invention, the transmission rate can be improved without increasing pilot frequency resources.

The invention discloses a port cargo real-time monitoring intelligent scene adaptation method based on sparse representation, and the method comprises the following steps: arranging a wireless signal transmitter and a receiver outside a cargo box, transmitting a wireless signal with a specific frequency to the cargo box by the signal transmitter, receiving a signal transmitted after the cargo passes through the signal receiver, and transmitting the received signal to the wireless signal receiver; according to the method, quantitative data of shielding effects of different types of goods on radio waves and labels of goods types are obtained, received signal features are analyzed, and matching with a pre-established goods type signal dictionary is carried out by utilizing a sparse representation technology so as to accurately identify the goods types. A dictionary of signal features representing different cargo categories is created, each cargo category corresponds to a specific signal mode in the dictionary, the system compares the received signals with the signal modes in the dictionary, and the specific category of the cargo is determined based on the comparison. According to the method, goods can be efficiently classified, manual box opening inspection in a traditional method is omitted, dependence on manual operation is reduced, high recognition precision and reliability are achieved, and the method is particularly suitable for logistics and transportation industries needing automatic recognition and classification of a large number of goods.

The application discloses a kind of AC / DC hybrid distribution network power flow calculation method and device considering incomplete LU decomposition preconditioning, and relates to the field of smart grid.The method comprises the following steps: establishing active / reactive control equation based on VSC steady-state equation;Based on Kirchhoff's current law, establish AC / DC node power imbalance equation;Taylor expansion is carried out on the above equation, and high-order terms are ignored, to obtain linearized unified power flow equation set and matrix form Jacobian matrix;Incomplete LU decomposition preconditioning is carried out on Jacobian matrix, and stable double-conjugate gradient method is used to solve, to obtain power flow solution.The application establishes unified linearized power flow model, combines incomplete LU decomposition preconditioning and stable double-conjugate gradient method, effectively reduces the sensitivity to initial value, improves the robustness and convergence speed of complex AC / DC hybrid system power flow calculation.

The present invention provides a method and system for quickly improving the activation sparsity of a large language model, including: step 1: modifying the activation function of the feedforward layer of the large language model to a ReLU activation function; step 2: adding a ReLU activation function after the output of a gated linear unit module for enhancing nonlinear capabilities in the feedforward layer of the large language model; step 3: cleaning and mixing pre-trained corpora from multiple sources; step 4: selecting optimal training hyperparameters according to the hyperparameter scaling law; step 5: training the model on a 200B corpus, and adopting a distributed training strategy based on data parallelism, model parallelism, and pipeline parallelism during the training process; step 6: fine-tuning the model in instruction alignment and human-preferred data sets. By introducing a higher quality and diversified corpus mixture, the present invention not only maintains the sparsity of the model, but also improves the overall capability of the model, making it applicable to a wider range of application scenarios.

The embodiment of the invention provides an information processing method and device, and the method comprises the steps: obtaining problem information corresponding to a natural language processing task, and inputting the problem information into a hybrid expert model; selecting a first expert model for the question information in the hybrid expert model, and determining a second expert model which is not selected in the hybrid expert model; constructing an embedded vector of the second expert model, and processing the embedded vector through a knowledge migration unit in the hybrid expert model to obtain a knowledge projection matrix; and updating the first expert model based on the knowledge projection matrix, and processing the question information by using the updated first expert model to obtain answer information.

The invention discloses a graph neural network enhanced combined personalized federal recommendation method, and belongs to the crossing field of federal learning and recommendation systems. The problem that the personalized recommendation effect is limited due to the fact that hidden information among users cannot be fully utilized due to privacy protection in an existing recommendation method is solved. The method comprises the following steps: constructing a user relation graph on a server by using locally updated project embedding without accessing a user interaction record; personalized items are embedded in different devices for local fine tuning, and global item embedding is obtained through calculation between user specific item embedding; the sparse global view is encouraged to save the communication cost of federated learning, two regularizers are used to ensure complementation of the two views, and finally, the weight is dynamically adjusted to realize combination personalization; a differential privacy technology is adopted, and privacy protection based on data locality is realized according to a data locality principle. According to the invention, on the premise of ensuring privacy security, the personalized performance and communication efficiency of the recommendation system are improved.

The invention belongs to the technical field of natural language processing and computer vision crossing, and discloses a multi-modal topic modeling method based on semantic consistency driving, which comprises the following steps: step 1, constructing a text-image multi-modal corpus and preprocessing; step 2, obtaining vector representation; step 3, creating a theme inference network; and step 4, training the topic inference network by taking the Dirichlet prior loss of the comparative learning loss, the multi-modal topic alignment loss and the energy loss on the multi-modal topic distribution as a joint optimization target, and completing cross-modal topic alignment and joint topic modeling. According to the method, the shared combined topic space is constructed, the topic distribution centrality of the same semantic samples is improved, the topic distribution overlapping degree of different semantic samples is reduced, it is ensured that the topic distribution of the text and the topic distribution of the image keep the corresponding relation statistically, and combined modeling and semantic association of the text and the image in the unified topic space are achieved.