Open AccessRobust Optimization of Convolution Natural Network
Author(s) -
Chongjie Ye
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1650/3/032105
Subject(s) - hyperparameter , computer science , artificial intelligence , machine learning , convolution (computer science) , noise (video) , hyperparameter optimization , optimization algorithm , deep learning , filter (signal processing) , algorithm , mathematical optimization , artificial neural network , computer vision , mathematics , image (mathematics) , support vector machine
Deep learning has played a very important role in computer vision. However, most of the methods used in computer vision highly rely on human to adjust the hyperparameter. That takes researchers lots of time, but the results sometime could not be most optimized. Besides, many architectures cannot perform robustly in training with noised data. This essay aims to solve the hyperparameter optimization problem by adapting the fruit fly optimization algorithm and suppose a high robust Convolution Natural Network including a Gaussian filter. Compared with methods such as FaceNe, InceptionV3 and Resnet5, GauCNN perform higher efficiency and accuracy with noise data.