Autoencoders architectures and applications

We will start with the basic, vanilla AE, then the deep version, followed by sparse autoencoders, denoising autoencoders, and wrap up with contractive autoencoders.

Throughout this entire section, we will use the credit card fraud dataset as an example to demonstrate how to apply autoencoders of various architectures.

Vanilla autoencoders

This is the most basic three-layer architecture, and is perfect to kick-start implementing autoencoders. Let's prepare the dataset. The dataset we are using is from a Kaggle competition and can be downloaded from the Data page in https://www.kaggle.com/mlg-ulb/creditcardfraud. Each row contains 31 fields, as follows:

• Time: Number of seconds...