Knowledge-Based Regularization in Generative Modeling

Naoya Takeishi, Yoshinobu Kawahara

Prior domain knowledge can greatly help to learn generative models. However, it is often too costly to hard-code prior knowledge as a specific model architecture, so we often have…