Rejection Sampling IMLE: Designing Priors for Better Few-Shot Image Synthesis
This addresses the challenge of training deep generative models with limited data for image synthesis applications.
The paper tackles the problem of poor test-time performance in few-shot image synthesis with IMLE-based methods by proposing RS-IMLE, which changes the prior distribution during training. This approach achieves substantially higher quality image generation compared to existing methods, as validated on nine few-shot image datasets.
An emerging area of research aims to learn deep generative models with limited training data. Prior generative models like GANs and diffusion models require a lot of data to perform well, and their performance degrades when they are trained on only a small amount of data. A recent technique called Implicit Maximum Likelihood Estimation (IMLE) has been adapted to the few-shot setting, achieving state-of-the-art performance. However, current IMLE-based approaches encounter challenges due to inadequate correspondence between the latent codes selected for training and those drawn during inference. This results in suboptimal test-time performance. We theoretically show a way to address this issue and propose RS-IMLE, a novel approach that changes the prior distribution used for training. This leads to substantially higher quality image generation compared to existing GAN and IMLE-based methods, as validated by comprehensive experiments conducted on nine few-shot image datasets.