Stochastic Optimization of Plain Convolutional Neural Networks with Simple methods

Convolutional neural networks have been achieving the best possible\naccuracies in many visual pattern classification problems. However, due to the\nmodel capacity required to capture such representations, they are often\noversensitive to overfitting and therefore require proper regularization to\ngeneralize well. In this paper, we present a combination of regularization\ntechniques which work together to get better performance, we built plain CNNs,\nand then we used data augmentation, dropout and customized early stopping\nfunction, we tested and evaluated these techniques by applying models on five\nfamous datasets, MNIST, CIFAR10, CIFAR100, SVHN, STL10, and we achieved three\nstate-of-the-art-of (MNIST, SVHN, STL10) and very high-Accuracy on the other\ntwo datasets.\n