The following are some of the practical applications where AEs may be used:

• Image coloring: Given a grayscale image as input, AEs can auto color the image and return the colored image as output.
• Noise removal: Denoising AEs are able to remove noise from images and reconstruct images without noise. Tasks such as watermark removal from videos and images can be accomplished.
• Dimensionality reduction: AEs represent the input data in a compressed form, but with a focus on key features alone. Therefore, things like images can be represented with reduced pixels, without much loss of information during image reconstruction.
• Image search: This is used to identify similar images based on a given input.

• Information retrieval: When retrieving information from a corpus, AEs may be used to group together all the documents that belong to a given input.
• Topic modeling: Variational AEs are used to approximate the posterior distribution, and it has become a promising alternative for inferring latent topic distributions of text documents.

We have covered the fundamentals that are needed for us to understand AEs and their applications. Let us understand, at a high level, the solution we are going to employ using AEs on the credit card fraud detection problem.