Improved Input Reprogramming for GAN Conditioning

TL;DR

This paper introduces InRep+, an improved input reprogramming algorithm for converting pretrained unconditional GANs into conditional GANs, especially effective with scarce, noisy, or imbalanced labeled data, outperforming existing methods.

Contribution

The paper identifies issues in existing input reprogramming methods and proposes InRep+ using invertible neural networks and PU learning to enhance performance in low-data scenarios.

Findings

InRep+ outperforms existing methods on CIFAR10 with 1% labeled data.

InRep+ achieves lower Intra-FID scores compared to previous approaches.

The method is robust to noisy and imbalanced label data.

Abstract

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.