Circuit structure for conducting least square equation solving according to positive definite symmetric matrices

Abstract

The invention belongs to the technical field of integrated circuit design, and particularly relates to a circuit structure for conducting least square equation solving according to positive definite symmetric matrices. The circuit structure is composed of an ACD computation module, an inversing module, a triangle multiplication module and a matrix multiplication module, wherein the ACD computation module is used for matrix decomposition, the inversing module is used for solving lower triangular matrix inverse matrices, and the triangle multiplication module and the matrix multiplication module are used for calculation of matrix multiplication. According to the circuit structure, the ACD decomposition algorithm is adopted for matrix decomposition, square root operation and repeated inversing operation are avoided, operation is simple, and the implementation area is small. Meanwhile, in an implementation process, the systolic array architecture is fully used and the multiplexing technique of the circuit structure is adopted, so that it is guaranteed that functions are correct, the utilization rate of hardware structures is increased, and the implementation area is reduced further. The circuit structure can be well used for solving least square equations.

Description

technical field [0001] The invention belongs to the technical field of integrated circuit design, and in particular relates to a circuit structure for solving a least square equation whose input is a positive definite symmetric matrix. Background technique [0002] The Nyquist criterion states that for a digital signal to be accurately recovered, the sampling rate must be greater than twice the signal bandwidth. With the development of science and technology and human society, people's demand for information is increasing day by day, and the frequency and bandwidth of the signal carried are also getting higher and higher, resulting in massive sampling data, which leads to huge demands on the sampling speed and storage of hardware. challenge. The recently proposed compressed sensing theory points out that when the signal satisfies sparsity, global observation can be performed at a speed much lower than the Nyquist sampling frequency, compression and sampling can be combined,...

AI Technical Summary

Problems solved by technology

For matrix inversion, it can be directly solved by the adjoint matrix, but as the size of the matrix increases, the hardware consumption of this direct solution algorithm will be very high

Images