Unsupervised Pre-Training of Deep Neural Classifiers

The paper is devoted to studying the issues of pre-training of deep neural network models. A generalized approach for pre-training deep models is proposed, which allows achieving better performance in final accuracy and can be used when training on small datasets. The effectiveness of the proposed approach is demonstrated on examples of training deep neural classifiers models using the MNIST and CIFAR-10 datasets. The obtained results can be used to train fully connected and convolutional types of layers.