Conditional Variational Autoencoders with Fuzzy Inference

We present an approach to constructing Conditional Variational Autoencoders (C-VAE) models with fuzzy inference during classification. This approach preserves the disentangling capabilities of the Variational Autoencoder (VAE) while simultaneously performing latent space clusterization. The Fuzzy C-VAE model provides useful features for anomaly detection, utilizing partially labeled datasets and controlled generation of new samples.