Estimation method for missing information of high-dimensional symmetric sparse network based on matrix decomposition

Abstract

The invention discloses an estimation method for missing information of high-dimensional symmetric sparse network based on matrix decomposition. The method comprises the following steps of: firstly, initializing a low-dimensional potential characteristic matrix, and determining the number of the low-dimensional potential characteristic matrixes and initialization numerical values of elements in the matrix; Designing an objective function based on known elements in the high-dimensional symmetric sparse network; Designing a solving algorithm of the objective function by utilizing a gradient learning method; Minimizing the objective function through a solving algorithm to obtain a potential characteristic matrix; Multiplying the potential feature matrix to obtain an estimation matrix of the high-dimensional symmetric sparse network, and obtaining missing information in the high-dimensional symmetric sparse network through the estimation matrix. According to the method, the defect that useful information is extracted from a high-dimensional sparse matrix in the prior art is overcome, missing information estimation accuracy and calculation efficiency are improved, and prediction symmetry and non-negativity are guaranteed to be met.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a method for estimating missing information in a high-dimensional symmetric sparse network based on matrix decomposition. Background technique [0002] With the advent of the Industry 4.0 era, many industrial applications, such as social service networks, e-commerce systems, bioinformatics applications, wireless sensor networks, etc., have also shown explosive growth in scale. In some cases, due to the difficulty of observing the relationships between the internal entities of these applications. Therefore, high-dimensional sparse matrix becomes a common form to describe this incomplete relationship. [0003] Although the high-dimensional sparse matrix is ​​sparse, it still contains a lot of useful information. For example, user preferences in recommender systems, protein connections in bioengineering, relative distances in wireless sensor networks, etc. At pres...

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved by the present invention is to provide a method for estimating missing information in high-dimensional symmetric sparse networks based on matrix decomposition. This method overcomes the existing defects of extracting useful information from high-dimensional sparse matrices, and improves the accuracy and accuracy of missing information estimation. Computationally efficient and guaranteed to satisfy predictive symmetry and non-negativity

Images