Graph structure searching method based on automatic machine learning

Abstract

The invention discloses a graph structure searching method based on automatic machine learning, which comprises the following steps: S1, sequentially extracting node information, node information andedges of a graph structure in each frame of image from an input image or video according to a time sequence to form a topological structure graph sequence Gt which is equal to (Vt, Et), wherein the edges in each frame of image form an edge set Et = {vtivtj (i, j) belongs to H}; and S2, constructing an adjacent matrix W = {Wij|i, j = 1,..., N} by using the edge set Et and the connection weight, initializing an adjacent matrix to obtain an initialized adjacent matrix W1; s3, obtaining a trained adjacent matrix W2, and obtaining a hyper-parameter adjacency matrix W3 from the adjacency matrix W2,wherein the hyper-parameter adjacency matrix W3 and the node set Vt form a high-robustness topological graph structure. A high-robustness graph structure is obtained by searching the whole sample setfor learning, and a weight matrix in graph convolution is expanded to the whole topological graph structure.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a graph structure search method based on automatic machine learning. Background technique [0002] With the advent of the era of artificial intelligence, the realization of intelligence in various fields has become the general trend, and it will also bring great convenience to people's lives. As an important branch of artificial intelligence, computer vision can be regarded as the "eyes" of artificial intelligence. Its main task is to use computers to analyze and process the collected information (pictures or videos) to understand the semantic information contained therein. [0003] Machine learning is an inevitable product of the development of artificial intelligence research to a certain stage. It is committed to improving the performance of the system itself by means of calculation and using experience. In computer systems, "experience" usually exists in ...

AI Technical Summary

Problems solved by technology

The input of these deep learning networks is a human-defined graph structure. For example, the input graph structure of the human behavior recognition network is based on the physical connection of human bones as the connection of the body skeleton topology, but human behavior is coherent, except In addition to the skeleton-based physical connection between bone nodes, there are also non-physical connections due to the coherence of actions. For example, when a person walks, in addition to the movement of the legs, it will also drive the swing of the arms. This information cannot be obtained from the physical connection of the skeleton. can get

Images