Method and system for adaptively setting deep belief network (DBN) parameter

Abstract

The invention discloses a method for adaptively setting a deep belief network (DBN) parameter based on a genetic algorithm. The method comprises the steps of setting an initial value for a DBN parameter to be optimized, and establishing a fitness function for setting the DBN parameter according to the initial value; obtaining optimal individual data by a genetic algorithm according to the fitness function; and anti-coding the optimal individual data to obtain an optimal DBN parameter. The problems that a manual parameter setting method is inefficient and often cannot achieve the optimal parameter setting are solved. The genetic algorithm can be used for improving the accuracy, the initial parameter of the DBN can be automatically determined according to input samples, and then the optimal network topology is obtained. The optimized parameter can be used for accurately learning the advanced characteristics of the sample data, so that the DBN can obtain better recognition results. The invention also discloses a system for adaptively setting the DBN parameter based on the genetic algorithm, which has the above-mentioned advantageous effects.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a method and system for adaptively setting parameters of a deep belief network. Background technique [0002] Currently, Deep Learning is a hot research topic in the field of object recognition. The concept of deep learning was proposed by Hinton et al. in 2006. It mainly simulates the learning process of the human brain through the neural network (Neural Network). neurons, and achieve the purpose of reconstructing objects through node weight training. Its most notable feature is that the feature extraction process reduces human intervention as little as possible. Deep learning is essentially a greedy algorithm, which is a neural network-like structure, but generally speaking, the network layer can have multiple layers, unlike artificial neural networks, which only have three layers. At the bottom layer of the deep learning model, that is, the input layer of the sample,...

AI Technical Summary

Problems solved by technology

In this way, since the parameters cannot be adaptively set according to the sample, it is difficult to obtain the optimal network parameters, and it is also impossible to obtain the optimal recognition effect.

Images