Controlling the Fidelity and Diversity of Deep Generative Models via Pseudo Density

TL;DR

This paper presents a novel method to control the fidelity and diversity of deep generative models by manipulating a new metric called pseudo density, enabling targeted adjustments through perturbation, importance sampling, and fine-tuning.

Contribution

It introduces pseudo density as a new metric and develops three techniques for biasing generative models towards higher fidelity or diversity.

Findings

Improved FID scores with minimal fine-tuning iterations

Effective per-sample adjustments for fidelity and diversity

Versatile techniques applicable to various deep generative models

Abstract

We introduce an approach to bias deep generative models, such as GANs and diffusion models, towards generating data with either enhanced fidelity or increased diversity. Our approach involves manipulating the distribution of training and generated data through a novel metric for individual samples, named pseudo density, which is based on the nearest-neighbor information from real samples. Our approach offers three distinct techniques to adjust the fidelity and diversity of deep generative models: 1) Per-sample perturbation, enabling precise adjustments for individual samples towards either more common or more unique characteristics; 2) Importance sampling during model inference to enhance either fidelity or diversity in the generated data; 3) Fine-tuning with importance sampling, which guides the generative model to learn an adjusted distribution, thus controlling fidelity and diversity. Furthermore, our fine-tuning method demonstrates the ability to improve the Frechet Inception Distance (FID) for pre-trained generative models with minimal iterations.