Multi-task sparse reconstruction and clustering method based on expectation propagation

Abstract

The invention discloses a multitask sparse reconstruction and clustering method based on expectation propagation, and the method comprises the steps: 1) constructing a Bayesian generation model with multitask sparse structure clustering through introducing 'spike and slab' and 'Dirichlet Process' prior probability distribution; (2) converting a model parameter posterior distribution solving problem into a parameter optimization problem of selected index class distribution by using an expected propagation technology; (3) initializing model approximate probability distribution; (4) respectively calculating the cavitity distribution of each q1j, q2j and q3j so as to update corresponding parameters, and stopping the operation when the model parameters converge to optimal values. According to the invention, the method can significantly reduce signal reconstruction errors, has stronger anti-noise capability, and can perform autonomous clustering on tasks based on sparse structure characteristics of signals without setting the number of categories in advance.

Description

technical field [0001] The invention belongs to the field of signal processing compressed sensing, and in particular relates to a multi-task sparse reconstruction and clustering method. Background technique [0002] Since the 20th century, sparsity has set off a huge wave in the field of signal reconstruction and compressed sensing, attracting countless researchers and engineers to invest in research one after another. Under the condition of signal sparseness, compressive sensing technology can reconstruct the original signal with high accuracy through fewer measurements. Especially in today's era of information explosion, sparse signal reconstruction can compress data and remove redundant information, which makes it widely used in radar imaging, radio astronomy, dictionary learning, image noise reduction, image classification and other fields. [0003] The method of sparse reconstruction is to reconstruct the signal using much less than the original measurement data. In r...

AI Technical Summary

Problems solved by technology

However, it is unrealistic in practical problems. In many cases, multiple signals often have several different sparse structures. At this time, still using all signal information to share the reconstructed signal will not only fail to improve the reconstruction accuracy, but will also lead to a sharp decline in performance.

Images