So far, in the previous chapters, we have learned several distance-based metric learning algorithms. We started off with siamese networks and saw how siamese networks learn to discriminate between two inputs, then we looked at prototypical networks and variants of prototypical networks, such as Gaussian prototypical networks and semi-prototypical networks. Going ahead, we explored interesting matching networks and relation networks.
In this chapter, we will learn about Memory-Augmented Neural Networks (MANN), which are used for one-shot learning. Before diving into MANN, we will learn about their predecessor, Neural Turing Machines (NTM). We will learn how NTMs make use of external memory for storing and retrieving information and we will also see how to use a NTM for perform copy tasks.
In this chapter, we will learn about the following:
- Reading and writing in NTM
- Addressing mechanisms
- Copy tasks using NTM
- Reading and writing in MANN