Neural network splitting learning algorithm

Abstract

The artificial neural network technology plays a more and more important role in the fields of mode recognition, image processing, automatic control, natural language processing, combinatorial optimization and the like through a specific information processing method and a network structure algorithm. In recent years, with the increase of data set samples, the number of layers of a deep learning network is more and more, and the calculation complexity is higher and higher. The accuracy of deep learning is improved by increasing the number of layers of a deep learning network, and the problem that the training speed of a neural network model is low occurs. The problems that the accuracy and the training speed of a deep learning network cannot be considered at the same time, and a deep learning network model cannot be reused are solved. According to the neural network splitting learning algorithm, a trained neural network function is split and supplemented, and a new neural network function is integrated. Only the supplementary part of the neural network function is trained, so that the training speed is increased, the accuracy of the original neural network function is improved, and the deep learning network model can be reused.

Description

technical field [0001] The invention relates to the field of artificial intelligence and machine learning, in particular to a neural network learning algorithm. Background technique [0002] With its unique information processing method and network structure algorithm, artificial neural network technology plays an increasingly important role in pattern recognition, image processing, automatic control, natural language processing, combinatorial optimization and other fields. In recent years, with the increase of data set samples, the number of deep learning network layers is increasing, and the computational complexity is getting higher and higher. By increasing the number of deep learning network layers to improve the accuracy of deep learning, the problem of slow neural network model training occurs. In order to solve the problem that the accuracy and training speed of the deep learning network cannot be balanced and the deep learning network model cannot be reused. The n...

AI Technical Summary

Problems solved by technology

[0003] For the accuracy of the deep learning network and the training speed can not be taken into account and the problem that the deep learning network model cannot be reused, the present invention is based on the basic neural network function For splitting, the neural network function will be supplemented Integrate into base neural network functions

Images