44 results about "Matrix addition" patented technology

The invention relates to tensor processing using low precision format. Aspects of the present invention are directed to computer-implemented techniques for improving the training of artificial neural networks using a reduced precision (e.g., float16) data format. Embodiments of the present invention rescale tensor values prior to performing matrix operations (such as matrix multiplication or matrix addition) to prevent overflow and underflow. To preserve accuracy throughout the performance of the matrix operations, the scale factors are defined using a novel data format to represent tensors, wherein a matrix is represented by the tuple X, where X=(a, v[.]), wherein a is a float scale factor and v[.] are scaled values stored in the float16 format. The value of any element X[i] according to this data format would be equal to a*v[i].

The present invention proposes a time-domain full-waveform inversion method based on amplitude incremental coding. By performing amplitude incremental coding on each sampling point of the simulated data and observation data and adding the amplitude polarity as a constraint, the same Data with different amplitude increments at time points; after encoding, construct a zero-setting matrix, multiply the matrix by the simulated data to set the data that causes cycle jumps to zero, thereby reducing the influence of this part of the data on the gradient; a part of the data After zeroing, the amplitude information of the original data is destroyed; in order to reduce the dependence of the inversion on the amplitude information and highlight the role of the phase information, a global cross-correlation objective function is used. The difference between this method and the traditional method lies in the encoding calculation and zeroing calculation of the simulated data and observation data, while the encoding calculation is simply matrix addition, subtraction, multiplication and division, which will not significantly increase the calculation time of the full waveform inversion. Comparing with the traditional full waveform inversion, the calculation efficiency is not reduced.

The invention discloses a large-scale MIMO signal detection method based on Jacobi iteration, and belongs to the technical field of wireless communication. The method comprises the steps: converting amatrix inversion process into an iteration process of matrix multiplication and matrix addition, providing the search direction for a Jacobi algorithm through a gradient algorithm and an overall correction acceleration method, and determining a correction coefficient of an iteration equation. According to the invention, the method performs the estimation of a high-dimensional matrix inversion process through an improved Jacobi iteration method, and converts the matrix inversion process into the iteration process of matrix multiplication and matrix addition, thereby greatly reducing the computing complexity. The gradient algorithm and the overall correction acceleration method are used for providing the search directions for the Jacobi algorithm and determining the correction coefficient of the iteration equation, thereby enabling the iteration convergence to be better, and enabling the convergence speed to be higher.

Belonging to the technical field of inertial navigation, the invention in particular relates to an inertial platform continuous roll self-calibration and self-alignment method under a static base. The method firstly adopts an inertial device input shaft as the reference to establish a systematic coordinate system, then on the basis of an inertial platform working principle, an inertial platform attitude angle is employed as the intermediate quantity to establish a system dynamics model and an observation model, and then a platform matrix addition scheme needed by inertial platform self-calibration and self-alignment is designed through observability analysis, finally the platform attitude angle and various error coefficients of the platform are selected as the state variables of the system, and autonomous calibration and alignment of the inertial platform can be realized through reduced cubature Kalman filter. The method provided by the invention can change the existing calibration and alignment modes of the inertial platform, simplifies the inertial platform self-calibration and self-alignment process, weakens the strong coupling between system calibration and alignment, and provides the basic theory and technical support for improving the inertial platform actual use precision.