Auto-Encoding Variational Bayes

This paper introduces Variational Autoencoders (VAEs), a deep learning framework for building probabilistic generative models. VAEs use neural networks to learn complex data distributions, allowing them to generate new data points similar to the training data. The key idea is to combine variational inference with neural networks to create a model that can both encode data into a lower-dimensional latent space and decode latent representations back into data space. This approach has been highly influential in generative modeling, enabling applications like image generation and data compression.