A method of using knowledge graph in zero-shot learning
A knowledge map and algorithm technology, applied in the field of zero-time learning, can solve problems such as inability to express "intimacy" and no class connection, etc., and achieve the effect of perfect description and good classification results
Active Publication Date: 2020-09-29
杭州淘艺数据技术有限公司
View PDF10 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, these semantic features are only one by one, and these categories are not well connected, and the degree of "intimacy" between these categories cannot be expressed intuitively.
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
[0029] The content of the present invention will be further described below in conjunction with the accompanying drawings.
[0030] Such as figure 1 Shown, the application method of the present invention, concrete steps are as follows:
[0031] Step (1) utilizes the ResNet deep convolutional neural network model training to obtain the visual features of the image;
[0032] Step (2) utilizes wordnet knowledge map to construct the relationship diagram between categories;
[0033] Use the wordnet knowledge map to construct a relationship diagram between categories in zero-shot learning. There are ancestors and descendants between categories. For example, tigers and lions belong to large cats, and tigers also have Siberian tigers and Sumatran tigers. According to these Relationships build a graph of ancestry between categories and descendant diagram
[0034] Step (3) calculates its weight relation according to the distance between nodes;
[0035] use Denotes the learned ...
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 provides a method for using a knowledge graph in zero-time learning. The present invention first utilizes the ResNet deep convolutional neural network model training to obtain the visual features of the image; then utilizes the wordnet knowledge map to construct a relational graph between categories; then calculates its weight relationship according to the distance between nodes; then utilizes the GraphSAGE algorithm to analyze the relational graph The nodes in the node are optimized; then the semantic features of the optimized class nodes are mapped to the same dimensional space as the visual features by using the graph convolutional neural network; finally, the category with the closest Euclidean distance to the visual features is found and used as the category for judgment. The present invention uses the knowledge map in the zero-time learning task, constructs the relationship diagram between the categories, adds more prior knowledge, utilizes the connection between the categories, and introduces the GraphSAGE algorithm, which can construct the relationship diagram The nodes in the node are optimized to make the description of the nodes more complete. The final classification results also have better performance.
Description
technical field [0001] The invention belongs to the technical field of zero-time learning, and the invention uses a knowledge map and a GraphSAGE algorithm on a zero-time learning task. Background technique [0002] In zero-shot learning, each category and its corresponding semantic features are given. The semantic features here include the attributes of the category, such as describing the size, color, etc. of these categories, and can also be the word vectors corresponding to these categories. However, these semantic features are only one by one, and these categories are not well connected, and the degree of "intimacy" between these categories cannot be expressed intuitively. The knowledge map just has the ability to integrate knowledge and connect knowledge. In addition, because the GraphSAGE algorithm can iteratively learn and aggregate neighbor node information, GraphSAGE can be used to optimize the node class in the relationship graph, so that the node representation ...
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 Patents(China)
IPC IPC(8): G06F16/36G06K9/62G06N3/04G06N3/08
CPCG06F16/367G06N3/08G06N3/045G06F18/22
Inventor 姜明刘志勇张旻汤景凡
Owner 杭州淘艺数据技术有限公司