NeurIPS 2021 Workshop

1. INTRODUCTION

Diffusion models have recently emerged as an expressive and flexible family of generative models, delivering competitive sample quality and likelihood scores on image and audio synthesis tasks

2. BACKGROUND

3. GUIDANCE

3.1 CLASSIFIER GUIDANCE

3.2 CLASSIFIER-FREE GUIDANCE

4. EXPERIMENTS

4.1 VARYING THE CLASSIFIER-FREE GUIDANCE STRENGTH

4.2 VARYING THE UNCONDITIONAL TRAINING PROBABILITY

4.3 VARYING THE NUMBER OF SAMPLING STEPS

5. DISCUSSION

6. CONCLUSION