A tensor compression method based on energy-gathered dictionary learning

Abstract

The invention discloses a tensor compression method based on energy gathering dictionary learning, and belongs to the field of signal processing. The method comprises the following steps: 1, respectively carrying out Tacker decomposition and sparse representation on tensor to obtain a dictionary, a sparse coefficient and a kernel tensor; 2, obtaining a new sparse representation form about the tensor according to the relationship between the sparse coefficient of the tensor and the kernel tensor; And 3, carrying out dimensionality reduction on the dictionary in the mapping matrix by utilizing an energy-gathered dictionary learning algorithm so as to realize tensor compression. According to the tensor compression algorithm based on energy gathering dictionary learning, tensor effective compression is achieved, compared with other compression algorithms, information of an original tensor can be more effectively reserved, and a better denoising effect is achieved.

Description

technical field [0001] The invention belongs to the field of signal processing, and in particular relates to a tensor signal compression algorithm based on energy-gathering dictionary learning, which can realize effective compression of the tensor signal. Background technique [0002] With the development of information technology, multidimensional signals play an increasingly important role in the field of signal processing. At the same time, the multidimensional (Multidimensional, MD) signal will also bring a great burden to the transmission and storage process. In order to deal with the challenges brought by multi-dimensional signal processing, the tensor representation of multi-dimensional signals has attracted people's attention. Representing multi-dimensional signals as tensors and processing them brings great convenience to the processing of multi-dimensional signals. Therefore, the essence of compressing multidimensional signals is to effectively compress tensors, s...

AI Technical Summary

Problems solved by technology

However, according to the famous "ugly duckling" theorem, there is no optimal schema representation without any prior knowledge, in other words, vectorization of tensor data is not always efficient

Specifically, this may lead to the following problems: First, destroying the inherent high-order structure and inherent correlation in the original data, information loss or masking redundant information and high-order dependencies in the original data, therefore, it is impossible to obtain a potentially more meaningful model representation; second, vectorization operations generate high-dimensional vectors, leading to "overfitting", "curse of dimensionality" and small sample problems

However, in the process of the tensor processing algorithm based on sparse representation, some new noises are often introduced, which will have a certain impact on the accuracy of the data

Moreover, in the process of sparse representation, the determination of sparsity will also bring certain challenges to the processing of tensors

Images