FractalNet

A FractalNet is a Deep Neural Network that is based on Self-Similarity Neural Network architecture.

• Context:
  • It was first developed by Larsson et al.,2017).
  • It is an alternative to ResNet.
  • …
• Example(s):
  • FractalNet: Ultra-Deep Neural Networks without Residuals (Larsson et al., 2017),
  • …
• Counter-Example(s):
  • DenseNet,
  • AlexNet,
  • GoogLeNet,
  • InceptionV3,
  • LeNet-5,
  • MatConvNet,
  • ResNet,
  • SqueezeNet,
  • VGG CNN.
• See: Deep Similarity Neural Network, Deep Convolutional Neural Network, Residual Network.

---

---

References

2017a

• (Huang et al., 2017) ⇒ Gao Huang, Zhuang Liu, Laurens van der Maaten, and Kilian Q. Weinberger. (2017). “Densely Connected Convolutional Networks.” In: Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2017). ISBN:978-1-5386-0457-1 doi:10.1109/CVPR.2017.243
  • QUOTE: FractalNets (Larsson et al., 2016) repeatedly combine several parallel layer sequences with different number of convolutional blocks to obtain a large nominal depth, while maintaining many short paths in the network. Although these different approaches vary in network topology and training procedure, they all share a key characteristic: they create short paths from early layers to later layers.

2017b

• (Larsson et al., 2017) ⇒ Gustav Larsson, Michael Maire, and Gregory Shakhnarovich. (2017). “FractalNet: Ultra-Deep Neural Networks Without Residuals.” In: Proceedings of the 5th International Conference on Learning Representations (ICLR 2017) Conference Track.

---