检查生成对抗网络判别器中的病态偏差：基于 StyleGAN3 模型的案例研究
Examining Pathological Bias in a Generative Adversarial Network Discriminator: A Case Study on a StyleGAN3 Model
AI总结：
本研究发现预训练 StyleGAN3-r 模型判别器存在无法用训练数据解释的内部颜色和亮度病态偏差，且该偏差导致评分按图像质量分层，对性别、种族等不同类别产生不成比例影响。
摘要：
生成对抗网络（GANs）生成逼真的人脸，人类通常难以将其与真实人脸区分。虽然机器学习模型中的偏差通常被认为是由训练数据中的偏差引起的，但我们在预训练的 StyleGAN3-r 模型的判别器中发现了病态的内部颜色和亮度偏差，这些偏差无法用训练数据来解释。我们还发现，判别器根据图像和人脸层面的质量系统地分层评分，这对不同性别、种族和其他类别的图像产生了不成比例的影响。我们考察了社会心理学刻板印象研究中常见的维度。
英文摘要：
Generative adversarial networks (GANs) generate photorealistic faces that are often indistinguishable by humans from real faces. While biases in machine learning models are often assumed to be due to biases in training data, we find pathological internal color and luminance biases in the discriminator of a pre-trained StyleGAN3-r model that are not explicable by the training data. We also find that the discriminator systematically stratifies scores by both image- and face-level qualities and that this disproportionately affects images across gender, race, and other categories. We examine axes common in research on stereotyping in social psychology.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
