Scholarlink

Abstract <p>We propose a progressive training strategy for quantum generative adversarial networks (QGANs) that enables stable image generation from 4 × 4 to 28 × 28 pixels via alpha-blended transitions. The method builds on the QINR-QGAN architecture with parameterized quantum circuits and data re-uploading. Experimental validation on MNIST demonstrates approximately twofold faster convergence to target-resolution quality while maintaining FID, SSIM, and PSNR metrics.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Progressive Hybrid Quantum–Classical Generative Adversarial Network for Image Generation

  • 期刊论文

  • 发表日期：

    2026年7月20日

查看全文
  • N. V. Ryabov

摘要

Abstract

We propose a progressive training strategy for quantum generative adversarial networks (QGANs) that enables stable image generation from 4 × 4 to 28 × 28 pixels via alpha-blended transitions. The method builds on the QINR-QGAN architecture with parameterized quantum circuits and data re-uploading. Experimental validation on MNIST demonstrates approximately twofold faster convergence to target-resolution quality while maintaining FID, SSIM, and PSNR metrics.