Attribute network representation learning method based on adaptive random walk
A random walk and learning method technology, applied in the field of network representation learning, can solve problems such as difficulty in learning the hierarchical information of attribute networks, failure to preserve long-tail distribution well, and ignoring the implicit relationship of attribute networks.
Inactive Publication Date: 2021-01-22
DALIAN UNIV OF TECH
View PDF0 Cites 4 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
First, existing work mainly focuses on the explicit relationships of nodes in topology information, ignoring the implicit relationships introduced by attributes in attribute networks.
Second, using this idea to directly perform random walks on the attribute graph cannot well preserve the long-tail distribution of vertices in the original attribute network
Finally, because the traditional method learns the interaction relationship between nodes in the Euclidean space, it is difficult to learn the network level information introduced by the attribute
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0036] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention.
[0037] The conceptual framework of the overall model is as figure 1 As shown, the general process is as follows: first, the attribute network is processed in terms of structure and attributes to obtain the corresponding network structure, and then a biased adaptive random walk is performed on the two networks, and the generated node sequence is input into the hyperbolic In the skip-gram model, node representations are learned.
[0038] In the first step, the explicit relation is sampled according to the attribute network structure information
[0039] In the construction of the construction structure diagram G s In the process of deleting the attribute nodes...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention belongs to the technical field of network representation learning, and provides an attribute network representation learning method based on adaptive random walk, which generates networknode representation according to the structure, attribute and hierarchical information of a network. The structural approximation of the network is reserved by sampling the topological structure of the attribute network; the semantic similarity of the network is reserved by capturing the high-order correlation of the node attribute bipartite graph network; then, hyperbolic spatial representationis learned by adopting a hyperbolic skimming algorithm, so that the complex hierarchical relationship of the attribute network is expressed by adopting a lower dimension; through modeling of multiplerelations of the network and adoption of a hyperbolic space learning method, the defect that implicit relations and hierarchical information are ignored in existing research can be remarkably overcome, and structural similarity, semantic similarity and hierarchical information in an attribute network are well reserved.
Description
technical field [0001] The invention belongs to the technical field of network representation learning, and relates to an attribute network representation learning method based on adaptive random walk. Background technique [0002] The network representation learning algorithm is responsible for learning the vector representation of each node in the network, and the learned node representation can be used as a feature of the node for subsequent network application tasks, such as node classification, link prediction tasks, etc. Networks are ubiquitous in real life, and the relationships between entities in fields such as paper citation relations, the World Wide Web, and online social networks can all be modeled as networks. However, in real systems, nodes are usually associated with a large amount of attribute data, which are usually complementary to the network. For example, in an academic network with abstracts of papers, papers with similar research methods often mention ...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06F16/35G06F16/36
CPCG06F16/353G06F16/367
Inventor 王宇新武彬张益嘉
Owner DALIAN UNIV OF TECH