Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning
arXiv e-Print archive - 2016 via Local Bibsonomy
This paper describes the CNN architecture Inception-v4.
They basically update Inception-v3 to use residual connections (see [He et al](http://www.shortscience.org/paper?bibtexKey=journals/corr/HeZRS15)). They also simplified the architecture as they moved from DistBelief to [TensorFlow](https://www.tensorflow.org/).
## Previous papers
* Inception-v1: [Going deeper with Convolutions](http://www.shortscience.org/paper?bibtexKey=journals/corr/SzegedyLJSRAEVR14)