Long short term memory architecture for working with sequences. Sequence to sequence model for example machine translation task and trick for faster training of decoder.

Method for getting simple contextual but interpretable word embeddings in which may conduct arithmetical operations and find distance between them.

FCN-similar network for image segmentation and also a few tricks which may improve segmentation and other CV solutions.

Article essence for FCN in which im introduced in segmentation and explain the fully convolutional networks needed for segmentation tasks.

New powerful blocks and techniques for ResNet which improve score and decrease count of parameters. New type of convolution.

It's pilot note. Simple and not bad image feature extractor network with low count of parameters that realise powerful idea of skip connections.