Hands-On Machine Learning for Algorithmic Trading

Convolutional autoencoders

In addition to feedforward architectures, autoencoders can also use convolutional layers to learn hierarchical feature representations. As discussed in Chapter 16, Deep Learning, feedforward architectures aren't well-suited to capturing local correlations typical of data with a grid-like structure.

Convolutional autoencoders, instead, leverage convolutions and parameter-sharing to learn hierarchical patterns and features irrespective of their location, translation, or changes in size.

We'll explore implementations of convolutional autoencoders for image data in the next section.