One Shot Learning using Memory-Augmented Neural Networks in Tensorflow.

Update: added support for Tensorflow v1*.

Tensorflow implementation of the paper One-shot Learning with Memory-Augmented Neural Networks.

Current Progress of Implementation:

•  Utility Functions:

•  Image Handler

•  Metrics (Accuracy)

•  Similarities (Cosine Similarity)

•  LSTM Controller and Memory Unit

•  Batch Generators

•  Omniglot Tester Code

•  Unsupervised Feature Learning through Autoencoders

•  Cattle/New Born Recognition

The benchmark dataset is Omniglot dataset. All the datasets should be placed in the data/ folder.

Adam Santoro, Sergey Bartunov, Matthew Botvinick, Daan Wierstra, Timothy Lillicrap, One-shot Learning with Memory-Augmented Neural Networks, [arXiv]