A fast semi-supervised classification method based on graph volume positive learning machine

Abstract

The present invention provides a fast semi-supervised classification method based on graph volume positive limit learning machine, comprising the following steps: constructing a self-expression model of disease classification data, and using the self-expression model to construct a global robust map of disease classification data, to obtain The adjacency matrix A of the disease classification data; calculate the random graph convolution model output H according to the adjacency matrix A; calculate the output layer weight β of the graph convolution limit learning machine according to the output H of the random graph convolution model; use the calculated The output layer weight β of the graph volume positive limit learning machine classifies the unmarked disease classification data; the beneficial effect of the present invention is: in the extreme learning machine method, the graph convolution network is introduced to replace the hidden layer, forming a brand-new graph Volumetric limit learning machine model; this model can handle non-European graph structure data, such as generalization to disease classification, bioinformatics, chemical medicine and other fields, while maintaining the fast learning speed and general approximation ability of extreme learning machines.

Description

technical field [0001] The invention relates to the fields of pattern recognition and data classification, in particular to a fast semi-supervised classification method based on a positive limit learning machine of graphs. Background technique [0002] Extreme learning machine (Extreme learning machine) is a very important technology, which has achieved great success in the fields of medical / biological data analysis, computer vision, image processing, and system modeling and prediction. Extreme learning machine is a special case of random vector functional-link network (Random vector functional-link network), which is a single hidden layer feed-forward neural network, in which the hidden layer is randomly generated, and the output weight value can be calculated Get an analytical solution. Since the extreme learning machine avoids the training of the hidden layer, this method has the advantages of small amount of calculation and fast operation speed. [0003] Although the e...

AI Technical Summary

Problems solved by technology

However, the disadvantage of this technique is that it needs to rely on gradient descent for optimization, the convergence speed is slow and it is easy to fall into local optimum

Images