Improved Input Reprogramming for GAN Conditioning
This addresses the GAN conditioning problem for scenarios with scarce, noisy, or imbalanced labeled data, offering an incremental improvement over prior input reprogramming methods.
The paper tackles the problem of converting a pretrained unconditional GAN into a conditional GAN with limited labeled data, proposing InRep+ which outperforms existing methods, achieving an Intra-FID of 76.24 compared to 114.51 for the second-best method on CIFAR10 with 1% labeled data.
We study the GAN conditioning problem, whose goal is to convert a pretrained unconditional GAN into a conditional GAN using labeled data. We first identify and analyze three approaches to this problem -- conditional GAN training from scratch, fine-tuning, and input reprogramming. Our analysis reveals that when the amount of labeled data is small, input reprogramming performs the best. Motivated by real-world scenarios with scarce labeled data, we focus on the input reprogramming approach and carefully analyze the existing algorithm. After identifying a few critical issues of the previous input reprogramming approach, we propose a new algorithm called InRep+. Our algorithm InRep+ addresses the existing issues with the novel uses of invertible neural networks and Positive-Unlabeled (PU) learning. Via extensive experiments, we show that InRep+ outperforms all existing methods, particularly when label information is scarce, noisy, and/or imbalanced. For instance, for the task of conditioning a CIFAR10 GAN with 1% labeled data, InRep+ achieves an average Intra-FID of 76.24, whereas the second-best method achieves 114.51.