Research on Stochastic Resonance of Convolutional Neural Network Based on Adam Optimization
SHANG Tianpeng
WANG Youguo
Abstract：In this paper,stochastic resonance is used to improve the performance of Adam optimized convolutional neural net-work under limited computational force.For the back-propagation algorithm,the momentum gradient descent algorithm is used to update the network parameters for Adam optimization,and the MNIST handwritten numeral set is used for simulation experiments.Under the experimental conditions of this paper,compared with the network of momentum gradient descent algorithm,the crossover entropy of Adam optimized network is reduced under the first 15 epochs.Increasing the number of training samples can reduce the reduction of cross entropy.This paper adds Gaussian noise to the output neurons of the convolution neural network optimized by Ad-am.The simulation results show that the stochastic resonance phenomenon occurs in the percentage of cross entropy reduction.In-creasing the number of training samples can reduce the effect of stochastic resonance phenomenon.
Keywords：gradient descent algorithmAdam optimizationGaussian noisesimulation experiment
Publication Date：2023-11-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：4( 2553-2556 )