A Generative Zero-Sample Prediction Method Based on Knowledge Graph

Abstract

The invention discloses a generative zero-sample prediction method based on a knowledge graph, comprising the following steps: taking hierarchically structured categories as category nodes, and connecting attribute descriptions, text descriptions and external knowledge as additional nodes for the categories to construct a fusion Knowledge graph of various semantic information; use graph neural network algorithm to encode semantic information of knowledge graph to generate category vector representation; use the generated category vector representation as the input of the generative model to generate samples of this category for zero-sample learning Algorithms for learning and prediction. By constructing a knowledge graph that integrates various semantic information, and based on the knowledge graph, samples with richer features and more discriminative between classes are generated for each invisible category, so as to better solve the prediction problem of unseen category samples.

Description

technical field [0001] The invention relates to the field of generative zero-sample learning, in particular to a generative zero-sample prediction method based on a knowledge map. Background technique [0002] Zero-shot Learning (ZSL) is an important branch in the field of transfer learning, which is mainly used to deal with the problem of missing samples in supervised learning. Typical supervised learning requires manual labeling of training samples to guide the machine learning model to extract features. The labeling of samples often requires huge manpower and financial resources, especially in classification problems. When some new categories appear, manual labor is required. Label hundreds or thousands of training samples. The heavy sample labeling work makes the model difficult to generalize. [0003] The zero-shot learning technology can deal with the learning and prediction problems of the model when the training samples are missing, that is, using the semantic prio...

AI Technical Summary

Problems solved by technology

However, there are many noises in these descriptions, and it is difficult to extract useful information

Images