Neural network training method based on Gaussian process prior guidance

Abstract

The invention discloses a neural network training method based on Gaussian process prior guidance. The neural network training method is used for improving the training process of a neural network toobtain a better training effect. The neural network training method specifically comprises the following steps: S1, obtaining a data set for neural network training, selecting a representative set formodeling priori knowledge, and defining an algorithm target; S2, performing a training process of first-stage batch iterative learning on the neural network model, and sequentially executing the steps S21 to S24 in each iterative batch; S3, after the current training process is finished, verifying the neural network model by using the verification set to obtain a verification set error rate of the current model; and S4, continuously repeating the steps S2 and S3 to perform a multi-stage training process on the neural network model until the model converges. According to the neural network training method based on Gaussian process prior guidance, the effectiveness of training can be effectively improved in tasks, and the network learning ability and the learning quality are improved, and the neural network training method has a good application value.

Description

technical field [0001] The invention belongs to the field of computer vision, in particular to a neural network training method guided by a Gaussian process prior. Background technique [0002] Image classification is the task of distinguishing different categories of images in a dataset. At present, the mainstream solution for image classification tasks is to train convolutional neural networks to solve the problem, and the training method generally uses the stochastic gradient descent method. In recent years, as the progress of the network structure has slowed down, the improvement of the training strategy has become increasingly important. Aiming at this goal, the present invention considers that in order to train a given model better in supervised learning such as image classification, it is necessary to provide as perfect and effective supervision information as possible. The dataset itself provides labels, but this inherent label only represents the classification re...

AI Technical Summary

Problems solved by technology

The data set itself provides labels, but this inherent label only represents the classification re

Images