Reusable and extensible machine learning method based on plug-in model

Abstract

The invention discloses a reusable and extensible machine learning method based on a plug-in model, which takes a lightweight feature learning model as a basic unit to construct a reusable and plug-in model to realize target part positioning and attribute analysis; semantic content necessary for a generative semantic description framework is provided for describing component features and component composition modes; time and space probability distribution of sample data is formed and serves as an important basis for target mode matching, so that intelligent analysis, learning and accurate and rapid identification tasks of a target are completed. According to the method, under the condition of limited samples and computing resources, the complexity of the model can be simplified, and the utilization efficiency of data samples is greatly improved.

Description

technical field [0001] The invention belongs to the technical field of machine learning, and in particular relates to a reusable and expandable machine learning method based on a plug-in model. Background technique [0002] At present, many mature "intelligent" model designs are essentially based on the category of "big data for small tasks". The complexity of data rules has brought great challenges to the specific task of generalization: small differences between data classes under the interference of "noise" can easily lead to false alarms in the pattern recognition system, while large differences within classes under numerous changes in data will lead to It is easy to cause identification failure. [0003] Therefore, models based on the "big data, small task paradigm" usually use massive data to generate and shape intelligent systems and models. However, the existing deep learning network model has a complex structure and a large number of parameters. The training and a...

AI Technical Summary

Problems solved by technology

[0004] In view of this, the present invention provides a reusable and scalable machine learning method based on a plug-in model, which can solve the generalization specific tasks of the existing "big data for small task paradigm (big data for small task)"

Images