Neural Network Topology Optimization Method Based on Three-way Decision

Abstract

The invention relates to a neural network topology optimization method based on three-branch decision-making, and aims at determining the number of hidden layer nodes of a neural network by an empirical formula method, which lacks certain theoretical support and the accuracy of the algorithm is not high. This is the first time that three-way decision theory has been applied to the problem of determining the number of nodes in the hidden layer of a neural network. First, initialize a hidden layer node, use the Focal loss loss function and the Adam algorithm to realize the learning process of the neural network; then, for the misclassified samples in the training phase of the neural network, use the three-branch decision theory, in the case of the minimum decision risk loss , divide it into different domains, and adopt corresponding strategies; finally, when the boundary domain is not an empty set, increase the number of hidden layer nodes of the neural network in turn, until the boundary domain is an empty set, stop the growth of the model, Therefore, the topological structure of the neural network is determined, and the prediction accuracy of the neural network is improved at the same time.

Description

technical field [0001] The invention belongs to the field of machine learning, and designs a neural network topology optimization method based on three-branch decision-making. The method can adaptively find the number of hidden layer nodes of the neural network, thereby realizing topology optimization. Background technique [0002] Neural network is one of the commonly used algorithms for machine learning, and it is a mathematical model that imitates the structure and function of biological neural networks. Network structure is the key to designing neural network algorithms, aiming to obtain the simplest possible structure, while enhancing the generalization ability of the network and improving the performance of the algorithm. For example, the single hidden layer feedforward neural network is the network with the simplest structure in the neural network and its derivative algorithms. It only consists of an input layer, a hidden layer and an output layer. The number of node...

AI Technical Summary

Problems solved by technology

In the growth neural network, considering the indistinguishable samples in the data set, the data set is divided only by increasing the number of nodes in the hidden layer or the number of layers in the hidden layer, which increases the complexity of the model on a large scale, and more importantly However, it reduces the efficiency of the model

Images