12 results about "Triangular decomposition" patented technology

The invention discloses an imitated radio frequency division multiplexing signal detection method based on low-complexity optimization expected propagation, and belongs to the signal detection technology of multi-carrier communication. According to the method, information is updated through iteration to approach an ML result, and the complexity is reduced from an exponential level to a cubic level; in order to reduce the complexity of the EP, sparse and quasi-banded channel structures in the DAFT domain are utilized. Through matrix block decomposition, upper and lower triangular decomposition and upper triangular matrix forward substitution, low-complexity matrix inversion is successfully realized, and the complexity of the EP algorithm is reduced to a linear level. In addition, the EP algorithm performance is further improved through a deep learning optimization moment matching strategy, so that the process becomes more adaptive through trainable positive and negative component parameters. In addition, DAFT domain iteration detection and decoding can be achieved based on the method, external information of a decoder is fed back to a detector, and the reliability of the whole system is remarkably improved. According to the invention, effective balance between performance improvement and calculation complexity is realized.

The application discloses a text reasoning method, product, equipment and storage medium, and relates to the technical field of computers, and comprises the following steps: through the orthogonal triangular decomposition mode and in combination with the judgment mechanism whether the pre-trained language model has been trained on the fine-tuning data set, if the pre-trained language model has been trained on the fine-tuning data set, the orthogonal triangular decomposition is performed on the to-be-fine-tuned parameter matrix, the triangular matrix obtained after the decomposition is fine-tuned and trained on the fine-tuning data set, and the fine-tuned language model is obtained; if the language model has not been trained on the fine-tuning data set, the orthogonal fine-tuning is performed on the language model starting from a preset unit matrix, the fine-tuned language model is obtained, and the text reasoning is performed by using the fine-tuned language model, through the above mode, the robustness of the language model to different data sets is improved, and the accuracy of the text reasoning result is improved, and meanwhile, the problem that the robustness of the traditional pre-trained language model parameter after fine-tuning is poor and the performance is reduced is solved.

The present invention discloses a motor drive control method based on a synchronous frequency back-EMF observer. The method is based on a control system comprising a back-EMF observer, a high-order phase-locked loop (PLL), a speed loop, a current loop, an SVPWM module, and an inverter, with the back-EMF observer connected to the high-order PLL. A computer storage medium is also disclosed. The present invention effectively solves the problem of inaccurate back-EMF information calculation, enabling real-time and accurate calculation of the rotor position of a permanent magnet synchronous motor (PMSM), improving the steady-state and dynamic performance of the PMSM drive control technology. The present invention constructs a synchronous frequency back-EMF observer based on a mathematical model of the PMSM voltage and a triangular decomposition of the calculated back-EMF signal. By utilizing its bandpass characteristics, the observer effectively suppresses harmonics and DC disturbances in the back-EMF signal, significantly improving the accuracy of rotor flux calculation.

A system encoder 300 reliably transmits a source data block (SDB) 311, configured for an outer transform matrix U and an inner transform matrix L. The inner encoder 301 receives the SDB and generates an output constraint block (OCB) 312, which is an image of the SDB under the inverse of a submatrix of the inner transform matrix. The outer encoder 302 receives a fixed data block (FDB) 313 and the OCB, and generates a transform output block (TOB) 315, which is an image of a transform input block (TIB) 314 under the outer transform matrix. The TIB transparently contains the FDB in its subblocks, and the TOB transparently contains the OCB in its subblocks. The inner encoder receives the TOB and generates a transmitted code block (TCB), which transparently contains the SDB in its subblocks.

This invention relates to the field of mine safety monitoring technology, specifically disclosing a method for deploying mine fire monitoring sensors. The method includes: establishing a three-dimensional model; acquiring fire smoke propagation data through numerical simulation to construct an original spatiotemporal concentration matrix; extracting threshold response time feature vectors and maximum time-varying gradient feature vectors from the original spatiotemporal concentration matrix; performing nonlinear logarithmic mapping and centering on the original spatiotemporal concentration matrix, and constructing a spatial feature basis matrix through singular value decomposition; concatenating the spatial feature basis matrix, threshold response time feature vectors, and maximum time-varying gradient feature vectors, introducing feature weight penalty coefficients, and constructing an augmented feature matrix; performing orthogonal triangular decomposition with column pivot selection on the transpose of the augmented feature matrix, and determining the optimal monitoring point based on the row index of the permutation matrix, thereby achieving effective monitoring and early warning of the initial stage of a fire. This improves the ability to capture low-concentration signals in the initial stage of a fire and the accuracy of reconstructing the overall concentration field.

The invention discloses a gesture recognition method and system, computer equipment and a readable storage medium, and relates to the technical field of image recognition. The method comprises the following steps: acquiring a to-be-recognized gesture image; and preprocessing the gesture image to be recognized according to the target detection algorithm, the single Gaussian model and the median filtering algorithm to obtain the preprocessed gesture image to be recognized. And performing feature extraction on the preprocessed to-be-recognized gesture image according to an orthogonal triangular decomposition algorithm and a tuck tensor decomposition algorithm to obtain a gesture feature set of the to-be-recognized gesture image. And inputting the gesture feature set of the gesture image to be recognized into the gesture recognition model to obtain a gesture recognition result. According to the method, the cyclic convolutional neural network can be improved, and the information of each signal mode is projected to a smaller subspace by using the full convolutional neural network, so that the method has higher accuracy in gesture recognition.

The present invention discloses a fast decomposition method for power flow calculation suitable for research purposes, and provides a Newton method for power flow calculation that is easy to modify and maintain for scientific researchers who conduct further research based on the fast decomposition method for power flow calculation. For an n-node power system, the node number is independent of the node type and can be arbitrarily numbered. The coefficient matrices B′ and B″ of the power flow calculation are stored as n×n. The elements of the rows corresponding to the balance node ΔP and the columns corresponding to Δθ in B′ are all 0, and the elements of the rows corresponding to the PV node or the balance node ΔQ and the columns corresponding to ΔU in B″ are also all 0, but the diagonal elements are not cleared to zero, and the original calculated values ​​are retained. In this way, the linear equation triangular decomposition method with excellent performance provided by the programming language can be used to solve the modified equation for power flow calculation, which reduces the programming difficulty, improves the calculation speed, and ensures the reliability of the algorithm.

This invention discloses a multi-core computing method and apparatus for QR decomposition that integrates column-block TSQR and tile update. The method includes: for a large-scale matrix to be decomposed, horizontally dividing it into blocks based on the number of cores in the multi-core processor, and further dividing the large-scale matrix into multiple layers according to the tile update method; constructing a multi-stage pipeline with a binary tree structure, performing parallel triangular decomposition on the initial layers of the large-scale matrix through the multi-core processor to obtain the corresponding upper triangular matrix and orthogonal matrix; replacing the original layer block matrix with the upper triangular matrix, performing conjugate transpose on the orthogonal matrix to obtain the update matrix, and implementing parallel updates for the remaining layers through the multi-core processor; based on the reflection vector implicitly stored by the multi-core processor during the parallel triangular decomposition process, solving for the overall orthogonal matrix corresponding to the large-scale matrix, and simultaneously outputting the overall upper triangular matrix obtained after layer update. This invention can realize QR decomposition of large-scale matrices.

The invention provides a method for preventing data leakage and related equipment, and relates to the field of artificial intelligence, and the method comprises the following steps: receiving a text input by a user; converting the text input by the user into one or more initial embedding vectors; executing encryption processing on the initial embedded vector in the process of executing the natural language processing, wherein the encryption processing comprises the following steps: generating a random matrix according to a random generator; determining a random orthogonal matrix according to the random matrix and an orthogonal triangular decomposition algorithm; performing orthogonal transformation on the initial embedded vector according to the random orthogonal matrix to obtain a transformed embedded vector; the transformed embedded vector is an encrypted vector. By means of the steps, accuracy and effectiveness of the data are guaranteed, and meanwhile safety of data information is improved.

A sparse sensor layout optimization method for turbine disk temperature field inversion, which obtains snapshot data of turbine disk temperature field evolution over time by finite element simulation at equal time intervals, and constructs it as structured tensor data, then constructs a reduced order model through high order tensor decomposition, subsequently obtains an optimized sensor layout through tensor fiber-pivoted orthogonal triangular decomposition, and finally realizes accurate reconstruction of turbine disk temperature field based on tensor compressive sensing method. The invention solves the problem of difficulty in realizing accurate turbine disk temperature field inversion using limited sparse sensor layout by capturing the physical information of turbine disk temperature field and constructing an accurate reduced order model, and has broad application prospects in the field of surface temperature field measurement of large rotating thermal mechanical hot end components.

The invention discloses a text reasoning method, a product, equipment and a storage medium, and relates to the technical field of computers, and the text reasoning method comprises the following steps: through an orthogonal triangular decomposition mode and in combination with a judgment mechanism for judging whether a pre-trained language model is trained on a fine tuning data set, if the pre-trained language model is trained on the fine tuning data set, judging whether the pre-trained language model is trained on the fine tuning data set; if yes, orthogonal triangular decomposition is carried out on the parameter matrix to be fine-tuned, fine-tuning training is carried out on a triangular matrix obtained after decomposition on a fine-tuning data set, and a fine-tuned language model is obtained; and if the language model is not trained on the fine tuning data set, performing orthogonal fine tuning on the language model by taking the preset unit matrix as a starting point to obtain a fine-tuned language model, and performing text reasoning by utilizing the fine-tuned language model. Therefore, the accuracy of a text reasoning result is improved, and meanwhile, the problems that the robustness is poor and the performance is reduced after parameters of a traditional pre-training language model are finely adjusted are solved.

The invention discloses a pre-training model low-rank adaptive parameter high-efficiency fine tuning method based on triangular decomposition, which comprises the following steps: for a pre-training large model, selecting levels needing parameter fine tuning from the large model according to actual needs, and respectively performing triangular decomposition on a weight matrix of each layer to be subjected to fine tuning, and taking the obtained scaling factor and the diagonal sub-matrix as trainable parameters, freezing other model parameters, inputting an input sample in the training sample set into the large model to update the trainable parameters, and reconstructing by adopting the updated trainable parameters to obtain a weight matrix, thereby completing efficient fine tuning of the pre-trained large model parameters. According to the method, the number of trainable parameters is remarkably reduced through three-solution decomposition, meanwhile, an orthogonal constraint mechanism is introduced, pre-training knowledge is kept, and meanwhile the performance level close to full-amount adjustment and optimization is achieved. The method not only reduces the consumption of computing resources, but also improves the optimization speed of the model.