Open AccessAn Analysis on Very Deep Convolutional Neural Networks: Problems and Solutions
Author(s) -
Tidor-Vlad Pricope
Publication year - 2021
Publication title -
studia universitatis babeş-bolyai. informatica
Language(s) - English
Resource type - Journals
eISSN - 2065-9601
pISSN - 1224-869X
DOI - 10.24193/subbi.2021.1.01
Subject(s) - computer science , normalization (sociology) , computation , convolutional neural network , residual , artificial intelligence , deep neural networks , deep learning , artificial neural network , algorithm , pattern recognition (psychology) , sociology , anthropology
Neural Networks have become a powerful tool in computer vision because of the recent breakthroughs in computation time and model architecture. Very deep models allow for better deciphering of the hidden patterns in the data; however, training them successfully is not a trivial problem, because of the notorious vanishing/exploding gradient problem. We illustrate this problem on VGG models, with 8 and 38 hidden layers, on the CIFAR100 image dataset, where we visualize how the gradients evolve during training. We explore known solutions to this problem like Batch Normalization (BatchNorm) or Residual Networks (ResNets), explaining the theory behind them. Our experiments show that the deeper model suffers from the vanishing gradient problem, but BatchNorm and ResNets do solve it. The employed solutions slighly improve the performance of shallower models as well, yet, the fixed deeper models outperform them.