In this chapter, we will discuss some neural models that can be employed for unsupervised tasks. The choice of neural networks (often deep ones) allows you to address the complexity of high-dimensional datasets with particular features that need sophisticated processing units (for example, images).
In particular, we will cover the following:
- Autoencoders
- Denoising autoencoders
- Sparse autoencoders
- Variational autoencoders
- PCA neural networks:
- Sanger's network
- Rubner-Tavan's network
- Unsupervised Deep Belief Networks (DBN)
