Use a VAE for Financial Data Augmentation

Contributed by: Sofien Kaabar

Difficulty level: Intermediate

Generative models are a class of machine learning tools designed to generate data that resembles a given dataset. These models learn the underlying patterns, statistical distributions, or structures within the data and can produce new samples that are statistically similar to the training data. Generative models have various applications, including image generation, text generation, and data augmentation.

A variational autoencoder (VAE) is a type of generative model. It is a probabilistic and generative model that extends the concept of traditional autoencoders by introducing probabilistic encoding and decoding. The way VAEs do their tasks is as follows:

1. The encoder takes an input data point and maps it to a probability distribution in a lower-dimensional latent space. The encoder can be considered as a recognition network. It consists of one or more neural network layers that encode the input data into mean and variance vectors representing the parameters of a multivariate Gaussian distribution in the latent space.
2. Data is then represented in what is known as the latent space. The latent space captures the underlying structure of the data. Instead of directly sampling from the Gaussian distribution in the latent space, the VAE samples from a standard Gaussian distribution (with mean 0 and variance 1) and then scales and shifts the samples using the mean and variance from ...