Imbalance-Robust and Sampling-Efficient Continuous Conditional GANs via Adaptive Vicinal Learning and Auxiliary Regularization

2025/08/03 by Xin Ding, Yun Chen, Ding, Xin +10
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Face recognition and analysis #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG) #Machine Learning in Healthcare

paper · pdf · doi:10.48550/arxiv.2508.01725

openalex publication_date 2025/08/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Recent advances in continuous conditional generative modeling, including Continuous conditional Generative Adversarial Network (CcGAN) and Continuous Conditional Diffusion Model (CCDM), estimate high-dimensional data distributions conditioned on scalar regression labels such as angles, ages, or temperatures. However, fixed-size vicinal training in CcGAN can be sensitive to non-uniform label densities, whereas CCDM relies on computationally expensive iterative sampling. To address these issues, we propose CcGAN-AVAR, an imbalance-aware extension of CcGAN that combines soft/hybrid adaptive vicinity with auxiliary discriminator-guided regularization. The adaptive vicinity constructs a label-dependent local radius according to the available samples around each target condition, and the multi-task discriminator supplies both a regression signal for label consistency and a density-ratio-estimation signal for distribution matching. We further provide a theoretical interpretation characterizing how adaptive vicinal weighting affects the local bias-variance behavior of the discriminator target, how hybrid truncation reduces objective-level cross-condition mixing, and how the density-ratio-based generator penalty approximates a Pearson Chi-square discrepancy up to the estimation error of the density-ratio branch. Extensive experiments on four datasets, including the newly constructed imbalanced RC-49-I, covering resolutions from 64x64 to 256x256 across eleven settings, demonstrate that CcGAN-AVAR obtains strong generation quality and label consistency while preserving the one-step sampling efficiency of GANs, achieving 300x--2000x faster inference than CCDM.

Citations

• Improved Distribution Matching Distillation for Fast Image Synthesis
• CCDM: Continuous Conditional Diffusion Models for Image Generation
• Analyzing and Improving the Training Dynamics of Diffusion Models
• Scaling up GANs for Text-to-Image Synthesis
• Scalable Diffusion Models with Transformers
• Rebooting ACGAN: Auxiliary Classifier GANs with Stable Training
• Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data
• Conditional GANs with Auxiliary Discriminative Classifier
• Efficient Subsampling of Realistic Images From GANs Conditional on a Class or a Continuous Variable
• Towards Real-World Blind Face Restoration with Generative Facial Prior
• Continuous Conditional Generative Adversarial Networks: Novel Empirical Losses and Label Input Mechanisms
• Denoising Diffusion Implicit Models
• Conditional Image Generation with One-Vs-All Classifier
• Differentiable Augmentation for Data-Efficient GAN Training
• DDcGAN: A Dual-Discriminator Conditional Generative Adversarial Network for Multi-Resolution Image Fusion
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Self-Attention Generative Adversarial Networks
• Group Normalization
• Spectral Normalization for Generative Adversarial Networks
• Mixed Precision Training
• Age Progression/Regression by Conditional Adversarial Autoencoder
• Conditional Image Synthesis With Auxiliary Classifier GANs
• f-GAN: Training Generative Neural Samplers using Variational Divergence\n Minimization
• Deep Residual Learning for Image Recognition
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Making a “Completely Blind” Image Quality Analyzer

Related

• CCDM: Continuous Conditional Diffusion Models for Image Generation
• Mining GOLD Samples for Conditional GANs
• Continuous Conditional Generative Adversarial Networks: Novel Empirical Losses and Label Input Mechanisms
• Continuous Conditional Generative Adversarial Networks (cGAN) with Generator Regularization
• S2cGAN: Semi-Supervised Training of Conditional GANs with Fewer Labels