d633: Capsule Networks Explained

Capsule Networks Explained – learn more | reddit discussion

Using a novel architecture that mimics the human vision system, Capsule Networks strives for translation equivariance instead of translation invariance, allowing it to generalize to a greater degree from different view points with less training data.

You can check out a barebone implementation of Capsule Network here, which is just a cleaned up version of gram.ai’s implementaion.

---

References

1. CNN: Translation Equivariance and Invariance
2. The Impact of Imbalanced Training Data for Convolutional Neural Networks
3. What is wrong with Convolutional Neural Networks?
4. Dynamic Routing Between Capsules